On June 2, 2015, President Barack Obama signed the USA FREEDOM Act of 2015, Public Law 114-23, a statute that ended a deeply controversial surveillance program and replaced it with something narrower, more supervised, and, within a few years, abandoned entirely. This guide carries a May 1, 2014 publication date, so every event described below happened after that date, and every one of them is dated explicitly. The story runs from the June 2013 disclosures that exposed the bulk telephone metadata program, through a federal appeals court ruling in May 2015, through a roughly one-day lapse in which major surveillance powers went dark, to the March 15, 2020 expiration of the underlying business records authority. The honest answer to the question this article exists to answer, whether bulk collection ended, is more interesting than either side’s version.
The Disclosure That Started Everything
On June 5, 2013, The Guardian published a classified order of the Foreign Intelligence Surveillance Court that required Verizon Business Network Services to turn over telephony metadata to the federal government on an “ongoing, daily basis.” The order was the first public proof of a program that had operated in secret for years: the daily, indiscriminate production of the telephone records of millions of people who were not suspected of any wrongdoing. The disclosures associated with Edward Snowden began on that date, and they continued through the summer of 2013, but the June 5 order was the document that forced the question Congress would spend the next two years answering.
The metadata at issue was transactional, not conversational. It included the numbers dialed and received, trunk and other routing identifiers, location data, and the time and duration of calls. It did not include the contents of any call, and it did not include subscriber names, addresses, or financial information. Defenders of the program would later stress these limits, and they mattered, but the privacy objection never depended on content. The objection was that the government was collecting, every day, a record of who called whom, when, and for how long, across the entire customer base of a major carrier, without suspecting any of those customers of anything.
The legal foundation claimed for this collection was the business records provision of the foreign intelligence statute, Section 215, codified at 50 U.S.C. 1861. That provision allowed the government to obtain tangible things, including business records, that were “relevant” to an authorized investigation, under 50 U.S.C. 1861(b)(2)(A). The government’s reading treated the word “relevant” as capacious enough to cover every telephone record in the country: because any single record might later prove connected to a terrorism investigation, the whole database was relevant in advance. That interpretation had been approved repeatedly by the surveillance court, which renewed the bulk production orders on a continuing basis, and it was unknown to almost everyone outside a small circle of executive branch officials, intelligence committee members, and the judges who signed the orders.
How did officials defend the bulk program as consistent with the statutory word relevant?
Officials argued that the entire telephone metadata database met the relevance test because any single record might later prove connected to a terrorism investigation, so collecting everything in advance was the only way to have the relevant records available when an investigation needed them.
The secrecy surrounding the interpretation was itself part of the controversy. A statute debated and passed in public was being read, in classified opinions, to authorize something the public text did not appear to authorize. Most members of Congress had no knowledge of the bulk program; the small circle that did know was bound by classification rules that prevented open debate. When the June 5, 2013 order appeared in print, lawmakers who had voted to reauthorize the business records provision learned, along with everyone else, what their votes had been taken to permit.
The political reaction unfolded over the next two years. The disclosures made the bulk program the central exhibit in a broader argument about the scale of post-2001 surveillance, and they created the constituency for reform that had not existed while the program was secret. The executive branch confirmed the program’s existence and defended its legality and value, while declassifying selected materials to make its case. Civil liberties organizations sued. Journalists and lawmakers demanded the release of the secret court opinions that had authorized the program. An earlier bill bearing the same short title was introduced in 2013, but it never reached the floor, and through 2014 competing reform proposals circulated without reaching the President’s desk.
The oversight board’s report, published January 23, 2014, added an official voice to the critics and is examined in detail later in this article. But legislation did not move until the sunset forced it to move. The pattern is familiar in the history of surveillance law: reform arrives not when the arguments are won but when the calendar leaves no alternative. The sunset of the business records authority, set for the end of May 2015, is what turned a debate into a deadline.
It is worth pausing on what the June 2013 order did and did not establish. It established that bulk collection was happening, that a major carrier had been ordered to participate daily, and that the surveillance court had approved the arrangement. It did not, by itself, establish whether the arrangement was lawful. That question would be answered, in part, by a federal appeals court less than a month before the sunset, and the answer would change the legislative dynamic in Washington.
The Business Records Power Before the Disclosures
To understand what the 2015 statute changed, it helps to understand what the business records power was before anyone outside a small circle knew how it was being used. Section 215 began as Section 215 of the 2001 antiterrorism law, which rewrote the business records provision of the foreign intelligence statute. Before 2001, the power was narrow, confined to particular categories of records sought in particular kinds of investigations. The 2001 law replaced those limits with a broad grant: the government could obtain “any tangible things,” including business records, upon showing that the things sought were relevant to an authorized investigation. The relevance standard, codified at 50 U.S.C. 1861(b)(2)(A), was the only substantive gate between the government’s request and the records of private parties.
Congress did not make the expanded power permanent. Like the roving wiretap authority and, later, the lone wolf provision, the business records power carried a sunset date, and Congress extended it several times, each time with a new expiration on the calendar. The sunset structure was a deliberate design choice: it forced periodic reconsideration of powers that would otherwise have become permanent features of the legal landscape without further debate. Each reauthorization was an opportunity for Congress to narrow, broaden, or simply extend the authorities, and each one passed with the understanding that the powers would come back for another vote.
What Congress did not know, during those reauthorizations, was how the executive branch and the surveillance court were reading the relevance standard. At some point after the 2001 expansion, the government began seeking, and the court began approving, orders that compelled the daily production of telephone metadata in bulk. The interpretation treated the entire body of domestic telephone records as “relevant” to counterterrorism investigations on the theory that any record might later prove connected to a suspect. The orders were classified, the legal reasoning appeared in classified opinions, and the circle of people who knew the full picture was small: executive branch officials, the judges who approved the orders, and a limited number of lawmakers with access to the classified briefings.
This is the context in which the June 2013 disclosures landed. The public debate that followed was not only about whether bulk collection was wise or lawful. It was about whether a democratic legislature can meaningfully oversee a power whose operative interpretation is secret. The reauthorizations Congress had passed were votes on a public text; the program the votes had sustained was governed by a classified reading of that text. The 2015 statute would address both problems, the collection itself and the secrecy that had sheltered it, but the secrecy problem had been building since the first bulk order was signed.
The Secret Legal Architecture
To understand why the 2015 statute took the shape it did, it helps to see exactly what the secret architecture looked like before the change. The government did not seize records from providers in the ordinary sense. It served them with court orders, renewed at regular intervals, compelling ongoing production of metadata that the companies already generated in the course of business. The providers complied under seal. The agency then stored the records in its own repositories, where analysts could query them under internal rules that required a terrorism related justification and, in theory, limited how far the queries could spread. The companies that handed over the data could not discuss the orders. The customers whose records were handed over had no way to know. The judges who approved the orders heard only the government’s side of the legal argument, because the surveillance court operated without an adversary in most proceedings.
This one sided procedure was the second target of the eventual reform, and it deserves emphasis because it explains why the bulk interpretation survived as long as it did. In an ordinary court, a creative reading of a statute faces an opposing lawyer whose job is to attack it. In the surveillance court, the government presented its relevance theory unopposed, year after year, and the judges who found it persuasive had no institutional mechanism for hearing the counterargument. The judges were not rubber stamps in any simple sense; the court’s opinions show serious engagement with the legal questions. But the structure of the proceeding meant that the strongest version of the opposing view was never presented by anyone with standing to present it. The 2015 statute would later address this defect directly, and that fix would prove more durable than the collection compromise the statute is usually remembered for.
The Appeals Court Removes a Pillar
On May 7, 2015, the United States Court of Appeals for the Second Circuit decided ACLU v. Clapper, 785 F.3d 787, and held that Section 215 had never authorized bulk collection. Judge Gerard Lynch, writing for the court, concluded that the word “relevant” in 50 U.S.C. 1861(b)(2)(A) could not bear the weight the government had placed on it. A relevance standard designed to let investigators obtain the records of a suspect could not be stretched to justify the indiscriminate collection of every telephone record in the country on the theory that any of them might someday matter. The full discussion of this ruling belongs with the broader set of surveillance law court cases that shaped the era, but its holding is the hinge on which the 2015 legislation turned.
The reasoning deserves attention because it was textual rather than policy-driven. The court examined what “relevant” means in the context of an authorized investigation and concluded that relevance must connect the records sought to the investigation at hand, not to some hypothetical future inquiry. A standard that treats everything as relevant because anything might become relevant is not a standard at all; it is a blank check. The government’s position, the court held, collapsed the distinction between records pertinent to an investigation and the universe of records in which pertinent ones might be found. That distinction was the only limit Congress had written into the provision, and the bulk program had erased it.
The court decided the case on statutory grounds and went no further. It declined to reach the First and Fourth Amendment claims the plaintiffs had raised, reasoning that the statutory holding resolved the dispute without requiring constitutional adjudication. It also declined to issue a preliminary injunction against the program, noting that the underlying authority was scheduled to expire within weeks and that the remedy, if any was needed, lay with Congress as it considered reauthorization. The restraint was deliberate: the court corrected the legal error and left the policy choice to the legislature.
Why did the appeals court decline to reach the constitutional claims?
The court resolved the case on statutory interpretation alone, holding that the statute never authorized the program, which made the constitutional questions unnecessary to decide, and it denied a preliminary injunction because the authority was about to expire within weeks, leaving the remedy to Congress.
The timing amplified the ruling’s force. With the sunset of the business records authority less than a month away, supporters of the existing program could no longer argue that Congress had already blessed bulk collection, because a federal appeals court had just held that Congress had done no such thing. The argument that the program rested on a settled legislative judgment collapsed. What remained was a choice: reauthorize the authority as it had been secretly interpreted, which the May 7, 2015 ruling had declared unlawful, or write a new framework that ended the bulk program and built something narrower in its place.
The ruling also vindicated, in legal terms, the position the government’s own oversight board had taken more than a year earlier. On January 23, 2014, the Privacy and Civil Liberties Oversight Board had concluded that the telephone records program “lacks a viable legal foundation under section 215,” a finding the Second Circuit’s statutory analysis echoed. The convergence mattered because it meant the legislative debate of May 2015 was no longer about whether bulk collection was wise; it was about whether it had ever been lawful, and the answer from both the oversight board and the appeals court was no.
There is a narrower lesson in the court’s restraint. By deciding on statutory grounds, the Second Circuit avoided creating constitutional precedent that might have constrained Congress’s options, and by withholding injunctive relief, it preserved the legislature’s freedom to design the replacement. The ruling cleared the legal ground without dictating the policy outcome. Congress would still have to decide what, if anything, should replace bulk collection, and it would have to decide under the pressure of an expiring clock.
The Word That Carried the Program
Every bulk collection regime in history has rested on a word, and the word here was relevant. Section 215 permitted the government to obtain tangible things, including business records, that were relevant to an authorized investigation. In ordinary legal usage, relevance is a filter: it separates the material connected to an inquiry from the material that is not. A subpoena for bank records relevant to a fraud investigation reaches the accounts tied to the suspected fraud, not every account in the banking system. The government’s theory inverted that function. It argued that the entire body of domestic telephone metadata was relevant to counterterrorism investigations generally, because any particular record might become relevant to some future investigation, and because the analytical method the agency favored required a complete dataset to work.
The logic had a surface plausibility that explains why judges accepted it more than once. Contact chaining, the technique at the center of the program, works by starting from a known suspect’s number and mapping outward through calling patterns. If the government holds only the records of people already under suspicion, chaining adds little; the value of the method lies in discovering previously unknown numbers that sit one or two steps from a target. A complete historical database lets analysts run that analysis backward in time the moment a new number becomes interesting, rather than starting collection from that moment forward. On this view, the whole dataset was relevant to every terrorism investigation in the same way that a library is relevant to every research project: not because each book bears on each question, but because the collection as a whole is the instrument through which questions get answered.
The flaw in the theory, as the Second Circuit explained on May 7, 2015, was that it read the limiting word out of the statute. If everything is relevant because anything might someday matter, then relevance imposes no limit at all, and Congress does not write limits that limit nothing. Judge Gerard Lynch’s opinion for the court treated the government’s reading as a category error: the statute authorized the collection of records relevant to an authorized investigation, which is to say a particular investigation, and the government had substituted a program of collecting all records for all possible investigations. The court decided the question on statutory grounds, declining to reach the constitutional claims, which gave the holding a particular kind of force. It did not say bulk collection violated the Constitution. It said Congress never authorized it, which is a narrower claim and, for legislative purposes, a more devastating one.
The statutory holding also explains why the surveillance court had approved the program for years without any judge publishing a dissent from the interpretation. The court’s judges were interpreting the same word, and in a proceeding where only the government argued, the government’s library theory arrived dressed in the language of deference to executive expertise in national security matters. The opinions approving bulk orders show judges working through the relevance question seriously and concluding, repeatedly, that the theory held. What none of those proceedings contained was the argument Judge Lynch eventually made: that a limit which limits nothing is not a limit. That argument required an adversary, or at least a judge hearing one, and the structure of the court supplied neither until the 2015 statute created the amicus mechanism.
There is a broader lesson in the career of that single word. Statutes governing secret programs are interpreted in secret, and interpretations adopted in secret tend to drift toward the needs of the interpreter. The relevance theory did not begin as a claim about collecting everything; it began as a sensible observation that investigative relevance is broader than trial relevance, and it grew, order by order, into a justification for a national database. Each renewal of the bulk order cited the prior renewals, and the weight of precedent accumulated without any external check. By the time the public learned of the program in June 2013, the interpretation had the solidity of settled law within the closed world that produced it, even though almost no one outside that world had ever examined it. The 2015 law’s answer was not to abandon the word relevant but to surround it with specific, judge tested constraints: the selection term, the dual showing, the hop limit, the time limit. Relevance remained the standard. It simply had to mean something again.
The Authorities Go Dark
Three surveillance powers shared a common sunset: Section 215, the business records provision; Section 206, the roving wiretap power codified at 50 U.S.C. 1805(c)(2)(B); and Section 6001 of the Intelligence Reform and Terrorism Prevention Act of 2004, the so-called lone wolf provision codified at 50 U.S.C. 1801(b)(1)(C). All three were scheduled to expire at 11:59 p.m. on May 31, 2015. The full history of the first of these powers is traced in the complete guide to the USA PATRIOT Act, which the 2015 statute would amend, but the essential point for this article is that the sunset was real, the date was fixed, and Congress had to act before it.
The Senate’s attempt to act failed on May 23, 2015. Cloture on H.R. 2048, the House-passed reform bill, fell short at 57 votes against the 60 required, after a filibuster by Senator Rand Paul that had consumed the available floor time. A separate proposal by Senator McConnell for a straight two-month reauthorization of the existing powers then failed 45 to 54. With no extension in place and the calendar exhausted, the three authorities expired at midnight as June 1, 2015 began. For roughly one day, the business records power, the roving wiretap power, and the lone wolf power were dark.
The lapse was brief, but it was genuine, and it remains one of the very few instances in this series in which a major surveillance authority actually went dark rather than being extended at the last moment. Section 215 reverted to its pre-2001 form, the narrower version of the power that had existed before the 2001 antiterrorism law expanded it. Grandfather clauses preserved the expiring powers for investigations that were already open, so ongoing counterterrorism and counterintelligence work continued under the old authorities as if the sunset had not occurred. What the lapse foreclosed was new reliance on the expired powers: no new bulk orders, no new roving wiretap applications, and no new lone wolf applications could be initiated during the gap.
What made the lapse a natural experiment rather than a mere pause?
The lapse created a rare before-and-after comparison: the same investigations, targets, and agencies operated under full powers, then under none, then under reformed powers, so the costs of the gap and the workability of the replacement could be observed instead of merely predicted.
On June 2, 2015, the Senate returned to H.R. 2048, rejected all three amendments offered by Senator McConnell, and passed the bill 67 to 32. President Barack Obama signed it that evening, restoring the three powers in modified form and ending the lapse. The restoration was not a simple reauthorization. Section 215 came back carrying the most significant structural change to domestic intelligence collection since the provision’s creation: a prohibition on the bulk collection the 2013 disclosures had revealed, and a new, narrower mechanism in its place.
The political meaning of the lapse deserves emphasis. The Senate had been warned for months that the sunset was coming, and it had a House-passed bill available. The failure to act before the deadline was not an oversight but the product of a genuine procedural fight, and the resulting gap demonstrated something legislators rarely get to see: what the absence of the powers actually looked like. The republic did not fall during the roughly one-day lapse. That observation would shape the debate over the replacement, because it weakened the claim that any interruption in collection authority spelled catastrophe.
The Senate’s Procedural Fight
The lapse did not happen because the Senate was unaware of the deadline. It happened because the chamber was genuinely divided over what should replace the expiring powers, and the division mapped onto a procedural stalemate that neither side could break in the time available. Understanding the fight requires understanding the three positions in play during May 2015.
The first position favored a clean reauthorization: extend the existing powers, including the business records provision as it had been interpreted, without structural change. Its supporters, led in the Senate by Senator McConnell, argued that the bulk program was lawful and valuable, that the Second Circuit’s May 7, 2015 ruling was wrongly decided or at least not final, and that the sunset deadline left no time for redesigning the collection architecture. The vehicle for this position was a straight two-month extension, which would have kept the powers alive while the debate continued. It failed 45 to 54 on May 23, 2015, well short of the votes needed.
The second position favored the House-passed reform bill, H.R. 2048, which ended bulk collection and created the provider-held replacement system. Its supporters argued that the Second Circuit had settled the legal question, that the oversight board’s January 2014 findings had settled the policy question, and that the only responsible course was to write the replacement into law before the sunset. The bill had passed the House 338 to 88 on May 13, 2015, and its supporters in the Senate believed a majority of the chamber favored it. But a majority is not always enough: on May 23, 2015, cloture on the bill failed 57 to 42, three votes short of the 60 required, after a filibuster by Senator Rand Paul.
The third position, held by Senator Paul and a small number of allies, was that neither the clean reauthorization nor the House bill went far enough, and that the authorities should be allowed to expire. Paul’s filibuster was the procedural mechanism that consumed the clock: by holding the floor, he denied the Senate the time it would have needed to resolve the contest between the first two positions before the May 31 deadline. The irony of the outcome is that the third position, which commanded the fewest votes, determined the result. The lapse was not the product of a majority decision to let the powers expire; it was the product of a minority’s ability to prevent a majority decision of any kind.
The sequence of the May 23, 2015 votes tells the story in miniature. First, the reform bill’s supporters could not break the filibuster, falling at 57 votes. Then the clean reauthorization’s supporters could not assemble a majority, falling at 45 votes. The chamber adjourned for the Memorial Day recess with no path forward and the clock running. When senators returned, the powers had already expired, and the politics had shifted: after the lapse, voting against the reform bill meant voting to keep the powers dark, rather than voting to preserve the old program. On June 2, 2015, the Senate passed H.R. 2048 by 67 to 32, rejecting all three McConnell amendments, and the President signed it that evening.
The episode is a case study in how sunsets shape legislative outcomes. Without the May 31, 2015 deadline, the bulk program’s supporters might have delayed reform indefinitely; with it, the reform bill’s supporters had leverage they would otherwise have lacked. The lapse itself, brief as it was, became evidence in the debate: the powers went dark, the grandfather clauses kept open investigations running, and nothing catastrophic occurred. That lived experience made it harder to argue, in the years that followed, that the reformed framework left the country defenseless.
How the Lapse Compares
Actual lapses of major intelligence authorities are rare, and the rarity is worth pausing over because it explains why the 2015 episode occupies a distinctive place in the series. Intelligence powers are typically reauthorized before their sunsets, often with little public attention, because the executive branch, the intelligence committees, and the congressional leadership share an interest in continuity. The sunset dates are known years in advance, the reauthorization bills are prepared in committee, and the votes are scheduled with room to spare. The system is designed to prevent exactly what happened in 2015.
The 2015 lapse occurred because the normal machinery broke down in the Senate. The House had done its part on May 13, passing the reform bill by an overwhelming margin. The Senate’s leadership preferred a different outcome and used the chamber’s procedures to pursue it, while the minority used the same procedures to block. The sunset, which normally functions as a forcing mechanism for timely action, instead functioned as a cliff that the Senate walked off. The result was a gap of roughly one day, contained by grandfather clauses and quickly repaired, but real nonetheless.
The comparison that illuminates the episode is not with other lapses, of which there are few, but with the near misses that characterize most sunset fights. In a typical reauthorization cycle, the deadline produces a flurry of negotiation and a last minute deal, and the authorities continue without interruption. The public sees the debate but not the darkness. In 2015, the public saw both, because the deal came a day late. That single day of darkness is what makes the episode analytically precious: it is the rare case in which the counterfactual, what happens without the authorities, was briefly actual. Most debates about intelligence powers argue about hypotheticals. For roughly one day in June 2015, the hypothetical was real, and the republic continued.
The Replacement: Records the Government Does Not Hold
The 2015 statute ended government-held bulk collection under Section 215 and replaced it with a fundamentally different architecture. Under Title I, Section 101, telephone providers retain their own call detail records in the ordinary course of business, and the government may obtain specified records only after securing the approval of the surveillance court for a specific selection term. The government no longer holds a database of everyone’s calls. It holds nothing until it asks, it asks only about a selector tied to a suspected foreign power or its agent, and a judge must approve the request before any records change hands.
The specific selection term is the load-bearing concept of the replacement design, and the statute defines it with care. Under 50 U.S.C. 1861(k)(4), added by Section 108 of the measure, a specific selection term must specifically identify a person, account, address, or personal device, or another specific identifier, and its use must be limited to the greatest extent reasonably practicable. The definition excludes the two evasions that would have recreated bulk collection under another name: a term may not be a broad geographic region, and it may not sweep in providers that are not subject to the order. This is the provision that answers the charge that the statute merely renamed the old program. A system in which the government must name its target, persuade a judge, and collect only from the providers that hold that target’s records is not the old system under a new label.
Approval requires two showings, and both must be made to the court. The FBI must demonstrate reasonable grounds to believe that the call detail records sought are relevant to an authorized investigation, and it must also demonstrate a reasonable, articulable suspicion that the specific selection term is associated with a foreign power or an agent of a foreign power engaged in international terrorism. The first showing preserves the traditional relevance standard; the second ties every production to a terrorism nexus. Production authorized under these standards is limited to 180 days and may be extended only by returning to the court, and the statute permits the government to obtain records two hops from the approved selector, meaning the contacts of the target and the contacts of those contacts, but no further.
To see how the design worked in practice, consider the sequence the statute required. An investigation identifies a telephone number associated with a suspected agent of a foreign power engaged in international terrorism. The FBI prepares an application naming that number as the specific selection term, setting out the reasonable grounds for relevance and the reasonable, articulable suspicion of the terrorism nexus. The surveillance court reviews the application and, if satisfied, issues an order directing the relevant providers to produce the call detail records associated with the term. The first production covers the records of the selector itself and its direct contacts, the first hop. A second, separately authorized step may reach the contacts of those contacts, the second hop. At every stage, minimization procedures require the prompt destruction of records that do not contain foreign intelligence information, so the material that accumulates is confined to what the investigation actually needs.
Minimization rules complete the design. The government must promptly destroy any records it obtains that do not contain foreign intelligence information, so irrelevant material cannot accumulate into a de facto database. The definition of call detail records performs the first round of minimization on its own: it excludes the contents of communications, the subscriber’s name, address, and financial information, and cell-site and GPS location data. What remains is transactional metadata, and even that may be retained and shared only under court-approved procedures. The contrast with the old program is structural, not rhetorical. The old program collected everything and asked questions later; the replacement requires the questions first, answered to a judge’s satisfaction, before any collection occurs.
Readers sometimes ask whether this authority overlaps with the separate Section 702 program for targeting non-U.S. persons abroad; it does not, and the explanation of FISA Section 702 describes the power this statute left alone. The 2015 measure’s one Section 702-related provision, Section 301, merely limited the government’s use of information derived from unlawfully certified targeting or minimization procedures, a narrow evidentiary rule rather than a reform of that separate program.
What the Statute Left Untouched
The 2015 measure was a targeted reform, and its limits are as important as its achievements. It amended the business records provision, the roving wiretap provision, and the lone wolf provision. It did not amend, restrict, or reauthorize Section 702 of the foreign intelligence statute, the authority added by the FISA Amendments Act of 2008 and codified at 50 U.S.C. 1881a, which permits the targeting of non-U.S. persons reasonably believed to be outside the United States. That program, including the practice of querying communications collected under it for information about Americans, sometimes called back door searches, continued under its own legal framework, unaffected by the 2015 reforms.
The distinction matters because the two authorities are frequently conflated in public discussion. Both involve the surveillance court, both involve the collection of communications-related information, and both were the subject of disclosures in 2013. But they rest on different statutory foundations, serve different purposes, and were treated differently by Congress. The 2015 statute’s reforms, the end of bulk collection, the specific selection term, the two-hop limit, the minimization rules, apply to the business records authority. Section 702 was left to its own separate reauthorization debates, on its own separate schedule.
Executive Order 12333 was likewise untouched. That executive order governs intelligence collection conducted overseas, outside the statutory framework of the foreign intelligence law, and the 2015 statute did not purport to regulate it. Any bulk collection conducted under that authority continued under the executive order’s own rules. This is the factual basis for the first of the two corrections this article makes: the claim that the 2015 statute ended all bulk collection is false, because the statute’s reach was confined to the business records provision, while other collection continued under Section 702 and Executive Order 12333.
The statute’s single Section 702-related provision deserves precise description, because it is easily overstated. Section 301 limited the government’s use of information derived from targeting or minimization procedures that had been unlawfully certified under Section 702. That is an evidentiary rule about the fruits of defective procedures, not a programmatic reform of the 702 program itself. It did not change who could be targeted, what could be collected, or how collected communications could be queried. Noting its existence is a matter of completeness; treating it as a 702 reform would be a misreading of what the provision does.
Two Changes to the Secret Court
The most durable achievements of the 2015 statute were not about collection at all. They were about the institution that approves collection: the Foreign Intelligence Surveillance Court, whose procedures were created by the 1978 foreign intelligence law described in the complete guide to FISA. For most of its history, that court had heard only one side. The government presented its applications, the judges ruled, and no one argued the other position. The 2015 measure changed that arrangement in two ways, and those two changes outlived everything else the statute built.
The first change was a panel of outside advisers. Section 401 of the statute, codified at 50 U.S.C. 1803(i) as Section 103(i) of the foreign intelligence law, requires the presiding judges of the surveillance court and its review court to jointly designate no fewer than five individuals eligible to serve as amicus curiae, friend of the court. The designees must possess expertise in privacy and civil liberties, intelligence collection, or communications technology, and they must be eligible for access to classified information. The operative command is mandatory in the cases that matter: the court shall appoint an amicus in any application or review presenting a novel or significant interpretation of the law, unless the court finds that appointment is not appropriate in that instance.
The design reflects a judgment about what had gone wrong. The bulk program had rested on a novel interpretation of the word “relevant,” adopted in secret opinions, without any advocate presenting the contrary reading. An amicus does not become a party to the case and does not receive a veto over the outcome, but the role guarantees that the judges hear an adversarial argument in precisely the cases where the law is being stretched. The appointment duty extends to the review court as well, so novel interpretations receive adversarial testing at both levels of the secret judiciary.
The one-sided character of the court’s proceedings had been a structural feature since the court’s creation, not an accident. In ordinary litigation, the adversarial process disciplines legal argument: each side’s claims are tested by an opponent with every incentive to expose their weaknesses. The surveillance court had no such discipline. The government was the only party in the room, and the judges, however conscientious, could only rule on the arguments presented to them. The amicus panel does not replicate full adversarial litigation, but it introduces a counterweight where none existed, and it does so at the exact point where the bulk program’s history showed the need was greatest: the interpretation of the law itself, rather than the facts of any individual application.
The appointment standard deserves close reading. The court must appoint an amicus when a case presents a “novel or significant interpretation of the law,” a threshold that captures precisely the kind of proceeding that produced the bulk program. Routine applications, renewals presenting no new legal question, do not trigger the duty. And the court retains an escape hatch: it may decline to appoint if it finds appointment inappropriate in the circumstances. The provision thus trusts the judges to identify the cases that need adversarial testing while ensuring that the default, in significant cases, is appointment rather than one-sided review.
The second change was a declassification review duty. Section 402 added Section 602 to the foreign intelligence law, codified at 50 U.S.C. 1872, directing the Director of National Intelligence, in consultation with the Attorney General, to conduct a declassification review of each decision, order, or opinion of the surveillance court or review court that includes a significant construction or interpretation of any provision of law. Each such opinion must be made publicly available to the greatest extent practicable. Redacted releases are permitted where national security requires withholding, and the duty may be waived for a particular opinion only with an unclassified summary explaining the basis for the waiver.
The target of this provision was the body of secret law that had grown up around the bulk program. Major legal interpretations, including the reading of “relevant” that sustained years of bulk orders, had bound the executive branch while remaining invisible to the public and to most of Congress. The declassification duty did not require the disclosure of sources, methods, or operational details; it required the disclosure of the law itself, on the theory that a legal interpretation secret from the public is difficult to square with the rule of law. Together with the amicus panel, it addressed the institutional conditions that had allowed the bulk program to persist: one-sided proceedings producing secret law.
Why did outside advisers matter more than the collection rules in the end?
The collection rules governed a program that lasted about four years before suspension, while the advisers and the declassification duty changed how the secret court operates in every significant case, so the institutional reforms kept working long after the program they accompanied had been shut down.
It is worth being precise about what these reforms did not do. They did not give the amicus the powers of a defense attorney, they did not open the court’s proceedings to the public, and they did not subject individual surveillance applications to adversarial challenge as a matter of course. The appointment duty applies to novel or significant legal interpretations, not to routine applications, and the court retains the power to find appointment inappropriate. The reforms were procedural rather than substantive, and their modesty is part of the explanation for their survival. They changed how the court thinks without dictating what it must conclude, and procedural changes of that kind proved harder to undo than the negotiated collection compromise that accompanied them.
The Strongest Case for the Program, and Against It
The debate over bulk collection produced two serious positions, and each deserves to be stated in its strongest form. The security argument, advanced by the executive branch in litigation and by the senators who sought a clean reauthorization of the existing powers, including Senator McConnell, ran as follows. Terrorist networks operate through chains of contacts, and the only way to map those chains quickly after identifying a suspect is to have the historical calling records already in hand. Telephone metadata reveals patterns, not conversations, and every production was approved by a federal court under a statutory standard. Speed matters in counterterrorism: when a new number surfaces in an investigation, analysts need to know immediately who that number has been calling, and a system that requires building the database after the fact surrenders the very advantage the program existed to provide. On this view, the program was a court-supervised tool for connecting known threats to unknown associates, and its bulk character was the source of its utility, not an accident.
The privacy argument, advanced by the ACLU as plaintiff in the litigation that reached the Second Circuit and by the government’s own oversight board, ran as follows. The program collected the telephone records of millions of people suspected of nothing, every day, on the theory that any of them might someday become interesting. That is suspicionless surveillance of the general population, and the fact that the records were metadata rather than content does not answer the objection, because metadata reveals associations, habits, movements, and relationships with a precision that content often cannot match. The program operated under a secret interpretation of a public statute, approved in one-sided proceedings, which meant the public could not evaluate, and Congress could not meaningfully oversee, what the law was taken to permit. On this view, the program traded the privacy of everyone for a speculative investigative convenience, and it did so without the democratic consent that a public reading of the statute would have required.
The oversight board’s findings on effectiveness, published January 23, 2014, cut against the security argument on its own terms. The Privacy and Civil Liberties Oversight Board concluded that the telephone records program “has shown only limited value” and reported that “we have not identified a single instance” in which it “made a concrete difference” in a counterterrorism investigation. The board found “no instance in which the program directly contributed to the discovery of a previously unknown terrorist plot or the disruption of a terrorist attack.” On the legal question, the board concluded the program “lacks a viable legal foundation under section 215” and found that it implicated First and Fourth Amendment concerns. By a 3 to 2 vote, the board recommended ending the program. The full assessment of these findings belongs with the analysis of the antiterrorism law’s civil liberties impact, but the essential facts are the date, the quotes, and the vote.
The 3 to 2 division on the board is itself informative. It shows that even among officials charged with overseeing civil liberties, the program had defenders as well as critics, and the recommendation to end it was contested rather than unanimous. The majority’s effectiveness findings, however, were stated as factual conclusions after review of the classified record: the board examined the investigations in which the program had supposedly played a role and found that, in each case, the same information had been or could have been obtained through other means. The claim was not that the program had never been useful in any degree, but that it had never made the concrete difference its defenders claimed, the discovery of an unknown plot or the disruption of an attack that would otherwise have succeeded.
Two misconceptions about the 2015 statute need direct correction, because each remains common. The first is the belief that the statute ended all bulk collection. It did not. The separate Section 702 program, which authorizes the targeting of non-U.S. persons abroad, continued under its own authority at 50 U.S.C. 1881a, and collection conducted overseas under Executive Order 12333 was never within the statute’s reach. What ended was the government’s bulk collection of domestic telephone metadata under the business records provision. The second misconception is the belief that the statute merely renamed the old program. It did not. Under the old program, the government held a bulk database and queried it at will; under the replacement, providers hold the records, the government holds nothing until it obtains court approval, and every production requires a specific selection term tied to a suspected foreign power or its agent. A system that requires naming the target and persuading a judge is not the old system under a new label, whatever its imperfections.
There is a final dimension to the debate that the effectiveness findings do not capture. Even if the bulk program had been more useful than the board found, its defenders would still have faced the legality question the Second Circuit answered in May 2015, and even if it had been lawful, its critics would still have faced the secrecy question the 2015 statute’s institutional reforms addressed. The debate operated on three levels at once: whether the program worked, whether it was legal, and whether a democracy can tolerate secret law. The 2015 statute answered the second question by ending the program and the third by opening the court, while the first question, effectiveness, was answered by the board’s January 2014 report and by the program’s own quiet abandonment in 2019.
The Legislative Vehicle and the Votes
The bill that became law was H.R. 2048, introduced on April 28, 2015, by Representative Sensenbrenner, a principal author of the original 2001 antiterrorism law who had become one of the bulk program’s most prominent critics within his own party. The House passed it on May 13, 2015, by 338 to 88, House Clerk Roll Call 224, a margin that reflected the unusual coalition behind reform. Senators Lee and Leahy introduced a Senate companion, S. 1123, but that bill was never voted on and did not become law; the Senate ultimately acted on the House vehicle directly.
The Senate’s proceedings in late May and early June 2015 were the most dramatic surveillance debate the chamber had seen in years. The May 23 cloture failure, the filibuster by Senator Rand Paul, the failed two-month clean reauthorization, and the roughly one-day lapse all preceded the final vote. On June 2, 2015, the Senate rejected all three amendments offered by Senator McConnell and passed H.R. 2048 by 67 to 32. President Barack Obama signed the bill that evening as Public Law 114-23, 129 Stat. 268. The title, in full, was the Uniting and Strengthening America by Fulfilling Rights and Ensuring Effective Discipline Over Monitoring Act of 2015, known universally as the USA FREEDOM Act of 2015.
The vote margins tell a story about the politics of the moment. A 338 to 88 House vote and a 67 to 32 Senate vote are not the margins of a partisan skirmish; they are the margins of a broad, bipartisan judgment that the bulk program could not continue. The coalition included civil liberties advocates who wanted the program abolished outright and security-minded legislators who accepted the replacement as the price of preserving the underlying authorities. That coalition held through the lapse and through the rejection of the McConnell amendments, which would have preserved more of the old framework. The statute that emerged was a compromise, but it was a compromise with a clear direction: away from government-held bulk collection, toward targeted, court-supervised production.
The Program Is Wound Down and the Authority Expires
The replacement program did not fail in Congress or in court. It failed in operation. In June 2018, the National Security Agency announced that it was purging three years of call detail records obtained under the 2015 statute because of technical irregularities. The agency had received records it was not authorized to receive: flaws in the production process had caused providers to deliver data beyond what the court’s orders permitted, and the agency could not reliably separate the properly obtained records from the over-collected ones. Rather than retain a database of uncertain legality, it deleted the entire three-year set. The purge was an admission that the machinery of the replacement, the handoffs between providers and the government that the statute’s design depended on, was not working as the court orders required.
The over-collection had a particular irony. The 2015 statute was written to end the indiscriminate acquisition of telephone records, yet the system built to replace bulk collection had itself acquired records beyond its legal authorization, through technical error rather than legal interpretation. The difference matters: the old program’s excess was a matter of law, a reading of “relevant” that the Second Circuit rejected, while the replacement’s excess was a matter of engineering, a failure to confine production to what the court had approved. But the practical result rhymed. Once again the government held telephone records it had no right to hold, and once again the remedy was deletion.
The consequence followed in early 2019. The agency suspended the call detail records program entirely and deleted the records it had collected under it, and the program was not renewed. The suspension was an administrative decision, not a statutory repeal: the legal authority for the provider-held system remained on the books, but the collecting agency concluded that the program’s operational difficulties outweighed its intelligence value. Less than four years after its creation, the substantive centerpiece of the 2015 statute had ceased to operate. The episode confirmed, in the hardest possible way, the article’s central observation about the difference between institutional reforms and collection compromises. The amicus panel kept functioning. The declassification duty kept functioning. The call detail records program did not.
The final chapter arrived on March 15, 2020. The business records authority, as reauthorized and reformed by the 2015 statute, expired on that date without reauthorization, and the roving wiretap and lone wolf provisions expired with it. The expiration was not a repeal of the 2015 measure; it was the sunset of the powers the measure had extended, arriving on the schedule Congress had written into the law. With the expiration, the legal foundation for any future call detail records program under the business records provision lapsed as well. The institutional reforms stood on separate footing: the amicus panel and the declassification review duty had been written into permanent sections of the foreign intelligence statute, not into the provisions that expired, so the court’s new procedures survived the sunset that ended the collection powers.
There is a temptation to read this ending as proof that the 2015 statute failed. That reading mistakes the nature of the achievement. The statute’s purpose was to end government-held bulk collection under the business records provision, and that collection never resumed. The replacement program’s abandonment does not resurrect the old program; it means that, for the years the replacement operated and after, the government held no bulk domestic telephone metadata database under this authority at all. Whether that outcome represents a triumph of reform or a quiet admission that the capability was never worth its costs depends on which of the two arguments above one finds more persuasive. What is not in doubt is the shape of the history: a bulk program exposed, a court ruling against it, a brief lapse, a replacement built and then abandoned, and an authority that expired, leaving behind the procedural reforms that had accompanied the whole experiment.
The Oversight Board’s Verdict in Detail
On January 23, 2014, the Privacy and Civil Liberties Oversight Board published its Report on the Telephone Records Program Conducted under Section 215, and the document landed like a verdict. The board was the oversight body charged with reviewing counterterrorism programs for their privacy and civil liberties implications, which made its conclusions difficult to dismiss as the work of outside critics. Here was an official entity, with access to the classified record, stating in plain language that the telephone records program was unlawful, of limited value, and should be ended. The report would become the evidentiary foundation for the reform position in the legislative debate of 2015, and its legal analysis would be echoed, more than a year later, by the Second Circuit.
The board’s legal conclusion was unsparing. The program, it found, “lacks a viable legal foundation under section 215.” The reasoning tracked the text of the provision: the relevance standard in 50 U.S.C. 1861(b)(2)(A) requires a connection between the records sought and an authorized investigation, and a program that collects every telephone record in the country cannot satisfy that requirement by asserting that any record might someday become relevant. The board’s reading of “relevant” anticipated, in substance, the holding Judge Gerard Lynch would issue on May 7, 2015: a standard that treats everything as relevant is not a standard at all. The convergence of the oversight board and the appeals court, arriving independently at the same textual conclusion, is what made the legal case against bulk collection so difficult to answer by the spring of 2015.
On the constitutional questions, the board found that the program implicated First and Fourth Amendment concerns. The First Amendment concern was associational: telephone metadata reveals the organizations people contact, the journalists they speak with, the religious and political groups they reach, and the systematic collection of such associations chills the freedoms of speech and association even when no content is captured. The Fourth Amendment concern was the suspicionless character of the collection: millions of people were subjected to ongoing government acquisition of their records without any individualized suspicion of wrongdoing. The board did not definitively resolve these constitutional questions, but its finding that they were implicated added weight to the statutory conclusion. A program of doubtful legality that also raised constitutional doubts was, in the board’s judgment, not a program the government should continue.
The effectiveness findings were the report’s most damaging section, because they addressed the program on its defenders’ own terms. The board examined the classified record, including the specific investigations in which government officials had said it had played a role, and concluded that the telephone records program “has shown only limited value.” The board reported that “we have not identified a single instance” in which the program “made a concrete difference” in a counterterrorism investigation. Most pointedly, it found “no instance in which the program directly contributed to the discovery of a previously unknown terrorist plot or the disruption of a terrorist attack.” In the cases where the program had contributed something, the board found, the same information had been or could have been obtained through other means. The claim was not that the program had never been useful in any degree, but that it had never delivered the decisive contribution its defenders described.
The board recommended ending the program by a 3 to 2 vote. The division matters for an honest account: the recommendation was contested, not unanimous, and the dissenting members did not share the majority’s conclusions. But the factual findings on effectiveness were stated as conclusions drawn from the classified record, and they stood as the most thorough official assessment available. The 3 to 2 margin also illustrated the broader shape of the debate: even among officials charged with protecting civil liberties, the program retained defenders, which is why the question had to be settled by legislation rather than by consensus.
The report’s afterlife confirmed its importance. Reformers in Congress cited its findings throughout the 2015 debate. The Second Circuit’s May 2015 decision echoed its legal analysis. And the report’s effectiveness conclusions supplied the factual premise for the article’s central claim: a program that had never made the concrete difference its defenders claimed was a program whose replacement, and eventual abandonment, cost less than its supporters had warned. The January 23, 2014 report did not end the bulk program by itself, but it made the program’s continuation intellectually untenable, which is often the necessary precondition for legislative action.
How the Replacement Changed Daily Practice
The difference between the old program and the replacement is easiest to grasp by comparing an ordinary day under each. Under the bulk program, the rhythm was standing and automatic: the surveillance court renewed the bulk production orders, the carriers delivered telephone metadata daily, the government accumulated it in its own database, and analysts queried that database as investigations required. Collection preceded suspicion. The database existed first, and the questions came later.
Under the replacement created by Title I of the June 2, 2015 statute, the rhythm is inverted. No standing database exists. Nothing is collected until an investigation identifies a specific selection term, a person, account, address, personal device, or other specific identifier that the statute’s definition, at 50 U.S.C. 1861(k)(4), requires to be drawn as narrowly as reasonably practicable. The FBI then prepares an application to the surveillance court, making two showings: reasonable grounds to believe the records sought are relevant to an authorized investigation, and a reasonable, articulable suspicion that the selection term is associated with a foreign power or an agent of a foreign power engaged in international terrorism. Only after the court approves does any provider produce any record.
The court’s role thus changed from wholesale to retail. Instead of approving a single bulk order that covered everyone’s records, the court reviews individual applications tied to individual selectors, each supported by its own factual showing. Production authorized under these standards is limited to 180 days, and an extension requires returning to the court with a renewed showing. The 180-day cycle builds periodic judicial review into the program’s operation: no production authority runs indefinitely on the strength of a single application.
The two-hop rule then governs how far the approved collection may reach. From the court-approved selector, the government may obtain the records of the selector’s direct contacts, the first hop, and the records of those contacts’ contacts, the second hop. No third hop is permitted. Each hop is bounded by the same court approval, the same relevance showing, and the same suspicion standard, and minimization procedures require the prompt destruction of any records that do not contain foreign intelligence information. The hop limit is the structural guarantee that a targeted query cannot expand, step by step, back into the indiscriminate collection the statute was written to end.
The definition of call detail records performs additional work at the definitional level. It excludes the contents of communications, so no conversation is captured. It excludes the subscriber’s name, address, and financial information, so the records are transactional rather than identifying in the first instance. And it excludes cell-site and GPS location data, so the physical movements of the people behind the numbers are not tracked through this authority. What remains is the metadata of calling relationships, and even that may be retained and disseminated only under court-approved procedures designed to limit both.
The providers, for their part, moved from the periphery of the system to its center. Under the old program, carriers delivered records in bulk and the government did the rest. Under the replacement, providers retain their own records in the ordinary course of business and must be able to isolate and produce the records tied to a court-approved selector within the court’s deadlines. The infrastructure burden of storage, retrieval, and compliance sits with industry rather than with the government. That shift was deliberate: it ensured that no government database of everyone’s calls would exist to be queried, even for legitimate purposes, because the database itself was the danger the statute was written to eliminate.
The New Procedures at Work
The amicus panel and the declassification review duty were designed to operate independently of any particular collection program, and that independence is the reason they survived the fate of the call detail records system. Understanding how they were built to work shows why procedural reforms outlast the substantive compromises they accompany.
The amicus design begins with designation. The presiding judges of the surveillance court and its review court jointly designate no fewer than five individuals eligible to serve as amicus curiae. The statute specifies the qualifications: expertise in privacy and civil liberties, intelligence collection, or communications technology, and eligibility for access to classified information. The expertise requirement ensures that the advisers can engage with the substance of the applications, which often turn on technical questions about how collection works and legal questions about how far the statute reaches. The classified-access requirement ensures that the adversarial testing happens on the full record, not on a redacted summary. An amicus who cannot see the classified materials cannot argue against the government’s position effectively.
The appointment duty is the heart of the provision. In any application or review presenting a novel or significant interpretation of the law, the court shall appoint an amicus unless it finds appointment inappropriate. The “novel or significant interpretation” threshold is aimed precisely at the kind of proceeding that produced the bulk program: not the routine renewal of an established authority, but the moment when the government asks the court to read the statute in a new way. The “inappropriate” exception preserves judicial discretion for the unusual case, but the default is appointment. The provision thus reverses the historical presumption of the court, from one-sided review as the norm to adversarial testing as the norm in the cases where the law is being developed.
The declassification review duty operates on a parallel track. The Director of National Intelligence, consulting with the Attorney General, must review each decision, order, or opinion of the surveillance court or review court that includes a significant construction or interpretation of any provision of law, and must make each such opinion publicly available to the greatest extent practicable. The “significant construction or interpretation” standard mirrors the amicus appointment threshold: both provisions target the moments when the secret court makes law, rather than the moments when it applies settled law to facts. Redacted releases are permitted where national security requires withholding particulars, and the duty may be waived for a specific opinion only with an unclassified summary explaining the basis for the waiver, so even the decision not to disclose must itself be explained in public.
Taken together, the two provisions address the two institutional conditions that had sheltered the bulk program: one-sided proceedings and secret law. The amicus ensures that the next novel interpretation will be argued against, not merely presented. The declassification duty ensures that the next significant interpretation will be visible, not merely filed. Neither provision dictates outcomes; both change the process by which outcomes are reached. And because they are written into permanent sections of the foreign intelligence statute, 50 U.S.C. 1803(i) and 50 U.S.C. 1872, rather than into the sunsetting collection authorities, they did not expire when the business records power expired on March 15, 2020.
The Two Courts
The 2015 reforms touched two courts, not one, and the distinction matters for understanding how the institutional changes were meant to work. The Foreign Intelligence Surveillance Court is the trial level tribunal that considers applications for surveillance orders, including the business records orders at the center of this article. The Foreign Intelligence Surveillance Court of Review is the appellate body that hears appeals from the trial court’s decisions. Both courts operate largely in secret, both hear the government’s arguments without an opposing party in the ordinary course, and both were addressed by the 2015 statute.
The amicus provision implicates both courts through its designation mechanism. The presiding judges of the two courts, acting jointly, must designate the no fewer than five individuals eligible to serve as outside advisers. The joint action requirement is a small but telling detail. It means neither court’s leadership can unilaterally shape the panel, and it ensures that the advisers enjoy the confidence of both the trial judges who will appoint them in initial applications and the appellate judges who may encounter them on review. The statute then provides for appointment in any application or review presenting a novel or significant interpretation of the law, which expressly includes appellate proceedings. The adversarial voice the reform introduced was thus available at both levels of the surveillance judiciary.
The declassification provision likewise reaches both courts. It requires review of each decision, order, or opinion of the surveillance courts, plural, that includes a significant construction or interpretation of any provision of law. The trial court’s bulk orders were the most consequential secret interpretations in the system’s history, but the review court’s rulings on novel legal questions are equally capable of shaping the law in secret. By covering both, Congress ensured that significant legal reasoning could not migrate to the appellate level to escape the sunlight requirement.
The two court structure also illuminates the limits of what the 2015 reforms attempted. The statute did not create a right of appeal for the targets of surveillance, who typically never learn they were surveilled. It did not open the courts’ proceedings to the public. It did not alter the standard of proof or the substantive requirements for most surveillance authorities. What it did was narrower and, in the drafters’ judgment, more achievable: ensure that when the courts make law, someone argues the other side, and ensure that when they construe statutes significantly, the public can eventually read what they said. Those are reforms to the conditions of decision rather than to the decisions themselves, which is why they belong to the institutional half of the statute and why they survived the collapse of the collection program they accompanied.
Why the Compromise Took This Shape
The 2015 statute was a compromise, and compromises take the shape of the coalitions that produce them. The coalition behind H.R. 2048 was unusual, and its composition explains both what the statute did and what it declined to do.
The bill’s House sponsor was Representative Sensenbrenner, who had been a principal author of the original 2001 antiterrorism law. His sponsorship mattered politically: criticism of the bulk program could not be dismissed as the work of perennial opponents of surveillance powers when it came from one of the authors of the powers themselves. Sensenbrenner introduced H.R. 2048 on April 28, 2015, and the House passed it on May 13, 2015, by 338 to 88, House Clerk Roll Call 224. A margin of that size, in a chamber often divided on security questions, reflected a genuine bipartisan judgment that the bulk program, as revealed and as ruled unlawful, could not continue.
The coalition united two groups that agreed on little else. Civil liberties advocates wanted the bulk program abolished and saw the bill as the vehicle for ending it. Security-minded legislators wanted the underlying authorities preserved and saw the bill as the price of reauthorization, the reform necessary to keep the business records, roving, and lone wolf powers alive past the May 31, 2015 sunset. The provider-held replacement was the point on which these groups converged: it ended the government database that the first group found intolerable while preserving a targeted collection capability that the second group considered necessary. A bill that had simply abolished the authority would have lost the second group; a bill that had reauthorized the old program would have lost the first. The specific selection term, the two-hop limit, and the court-approval requirement were the technical expression of that political convergence.
The Senate’s endgame confirmed the coalition’s breadth. On June 2, 2015, the Senate rejected all three amendments offered by Senator McConnell and passed H.R. 2048 by 67 to 32, a margin that, like the House vote, signaled broad rather than narrow support. The Senate companion bill, S. 1123, introduced by Senators Lee and Leahy, was never voted on and did not become law; the Senate worked directly from the House vehicle, which is why the enacted statute is H.R. 2048 in its House-passed form as amended by the Senate’s acceptance. President Barack Obama signed the bill that evening as Public Law 114-23.
The sunset deserves credit as the forcing mechanism. Without the May 31, 2015 deadline, the bulk program’s defenders might have delayed reform indefinitely, and the unusual coalition might never have been tested. The roughly one-day lapse that resulted from the Senate’s procedural fight became, paradoxically, an asset for the reformers: it demonstrated that the powers could go dark without catastrophe, which weakened the argument that only the old program could keep the country safe. Deadlines concentrate minds, and this deadline concentrated them on the compromise that became law.
What the 2020 Expiration Did and Did Not End
On March 15, 2020, the business records authority expired without reauthorization, and the roving wiretap and lone wolf provisions expired with it. The expiration is sometimes misdescribed as a repeal of the 2015 statute. It was not. It was the sunset of the powers the statute had extended, arriving on the schedule Congress had written into the law. The distinction matters: a repeal removes a law from the books, while a sunset lets specified provisions lapse while leaving the rest of the statute intact. The 2015 measure’s collection authorities lapsed; its institutional reforms did not.
By the time of the expiration, the collection program the statute had created was already gone. The National Security Agency had suspended the call detail records program in early 2019, deleted the records it had collected, and declined to renew it. The June 2018 purge of three years of over-collected records had foreshadowed the end. So the March 2020 sunset did not kill an operating program; it removed the legal foundation on which a future program might have been built. The underlying business records power, the authority that had sustained first the bulk program and then its replacement, ceased to exist as a basis for new collection.
What survived was everything the statute had written into permanent law. The amicus panel, codified at 50 U.S.C. 1803(i), sits in the section of the foreign intelligence statute governing the court’s procedures, not in the sunsetting collection authorities. The declassification review duty, codified at 50 U.S.C. 1872, likewise stands apart from the expired powers. Both provisions continued to operate after March 15, 2020, because neither depended on the business records authority for its legal existence. This is the structural reason the article’s central claim holds: Congress placed the institutional reforms where sunsets could not reach them, whether by design or by the logic of drafting procedural provisions into procedural sections.
The authorities the 2015 statute never touched were, of course, unaffected by the expiration. Section 702, the separate program for targeting non-U.S. persons abroad under 50 U.S.C. 1881a, continued under its own authorization and its own schedule. Collection under Executive Order 12333 continued under that order’s framework. The 2020 expiration ended the business records, roving, and lone wolf powers and nothing else. The resulting landscape is the one the table below summarizes: no bulk telephone metadata program under the business records provision, no replacement call detail records program in operation, and a surveillance court that hears adversarial argument in significant cases and publishes its significant legal interpretations to the greatest extent practicable under the permanent procedural reforms. The experiment that began with the June 2013 disclosures ended, five years after the statute and seven years after the disclosure, with the procedures as its legacy.
The Two Showings the FBI Must Make
The replacement program’s most important safeguard is also its most technical: before the government may obtain any call detail records, the FBI must satisfy the surveillance court on two separate points, and both must be established in every application. The dual burden is what separates the new system from the old one, where a single, capacious reading of relevance was enough to sustain the collection of everything. Understanding the two showings is understanding why the replacement could not recreate the bulk program even if the executive branch had wanted it to.
The first showing is familiar: reasonable grounds to believe that the call detail records sought are relevant to an authorized investigation. This is the traditional relevance standard, the same concept that appeared in the pre-2015 version of the provision, but confined under the new design to records tied to a specific selection term rather than to the universe of records. The government must connect the records it wants to an investigation that is itself authorized, meaning an investigation conducted under the guidelines and authorities that govern foreign intelligence work. Relevance, in this setting, is a relationship between particular records and a particular inquiry, not a property of the entire national telephone system.
The second showing is the innovation: a reasonable, articulable suspicion that the specific selection term is associated with a foreign power or an agent of a foreign power engaged in international terrorism. This is a higher and more particularized burden than relevance. It requires the FBI to articulate facts, not hunches, connecting the selector, the telephone number, account, or device, to a foreign power or its agent, and further to international terrorism. The suspicion must be reasonable, meaning a judge must find it objectively justified, and it must be articulable, meaning the FBI must state it in the application rather than merely assert it. A selector that cannot be tied to international terrorism cannot be approved, no matter how relevant its records might be to some other investigation.
The terrorism nexus is doing critical work in the design. The old bulk program collected the records of everyone, including people with no conceivable connection to terrorism, on the theory that any record might become relevant. The new program permits collection only when the government can already articulate a terrorism-related suspicion about the selector. The suspicion requirement thus functions as a gate at the front of the process: it ensures that collection begins with a reason to look at a particular target, rather than with the indiscriminate acquisition of records in the hope that a reason will emerge later. This is the structural inversion the statute intended.
The two showings also interact. Relevance without the terrorism nexus is insufficient, and the terrorism nexus without relevance to an authorized investigation is insufficient. An application that demonstrates a strong terrorism connection but seeks records unrelated to any authorized investigation must be denied. An application that demonstrates relevance to an investigation but cannot articulate the terrorism nexus must be denied. The court reviews both, and the statute gives the court no authority to waive either. The conjunction is what makes the standard meaningful: each showing constrains a different dimension of the government’s request, and together they confine collection to the intersection of investigative need and terrorism-related suspicion.
The standard of proof deserves attention as well. “Reasonable grounds” and “reasonable, articulable suspicion” are both below probable cause, the standard for a criminal search warrant, and the choice was deliberate. Congress did not require the government to meet the criminal standard before obtaining telephone metadata in a foreign intelligence investigation; it required a showing calibrated to the intelligence context, where the purpose is prevention and the records sought are transactional rather than content-bearing. Critics of the program argued that the standard was too low, while defenders argued that anything higher would have made the replacement unworkable. The statute resolved the dispute in favor of the intermediate standard, and the court’s role is to enforce it as written.
The 180-day limit on production adds a temporal dimension to the showings. An approval does not authorize collection indefinitely; it authorizes production for 180 days, after which the government must return to the court for an extension. Each extension requires a renewed application, which means the two showings must be made again with current facts. A selector whose terrorism nexus has dissipated, or whose records are no longer relevant to an authorized investigation, cannot be extended on the strength of the original application. The renewal requirement prevents the slow drift by which a targeted authority can expand, over time, into something broader than its original justification.
The minimization rules then govern what happens to the records once they are obtained. The government must promptly destroy any records that do not contain foreign intelligence information, so the fact of lawful collection does not become a license for indefinite retention. The destruction duty applies continuously, not just at the end of the 180-day period: as analysts review the produced records, the ones that prove irrelevant must be purged. Combined with the definitional exclusions in the call detail records definition, no content, no subscriber identifying information, no location data, the minimization regime is designed to ensure that what the government keeps is confined to what the investigation needs.
The two-hop rule operates within this framework as a further constraint. Even with a properly approved selector, the government may reach only two steps outward through the calling network: the selector’s direct contacts, and the contacts of those contacts. The hops are not independent grants of authority; they are bounded extensions of the original approval, subject to the same court oversight and the same minimization duties. The hop limit recognizes a practical reality of investigations, that the associates of a suspect matter, while refusing to let that reality become a justification for unbounded network mapping. Two hops from a terrorism suspect is an investigation; unlimited hops would be the bulk program reconstructed one selector at a time.
It is worth contrasting this architecture with the one it replaced. Under the bulk program, the government made a single relevance argument, once, about all records, and the court’s approval sustained ongoing daily collection. Under the replacement, the government must make two showings, per selector, per 180-day period, to obtain only the records tied to that selector and its two-hop neighborhood. The administrative burden of the new system is vastly greater, and that burden is the point. A system that is easy to operate at scale invites operation at scale; a system that requires individualized judicial approval for each target resists it. The friction is the safeguard.
The Other Two Powers: Roving Wiretaps and the Lone Wolf
The 2015 debate is usually remembered as a debate about bulk telephone metadata, but the statute reauthorized three powers, not one, and the other two deserve their own account. The roving wiretap authority and the lone wolf provision shared the May 31, 2015 sunset, the roughly one-day lapse, and the March 15, 2020 expiration. Their histories explain why Congress bundled them with the business records provision and why their fate was tied to the bulk collection fight even though neither involved bulk collection.
The roving wiretap authority, Section 206 of the 2001 antiterrorism law codified at 50 U.S.C. 1805(c)(2)(B), addresses a practical problem in surveillance: targets change phones. Under traditional wiretap law, a surveillance order specifies the facility to be monitored, a particular telephone line or communications account, and investigators must return to court for a new order each time the target switches to a new facility. A sophisticated target can defeat surveillance simply by discarding phones faster than the government can obtain orders. The roving provision solves this by letting the order follow the person rather than the facility: once the court finds that the target’s actions may thwart surveillance, the order authorizes monitoring across the facilities the target uses, without a new application for each one.
The authority was less controversial than the bulk program, but it was not uncontroversial. Civil liberties advocates noted that roving orders necessarily involve monitoring facilities that may be used by people other than the target, since a phone the target picks up may also be used by others, and the particularity that traditionally constrains searches is relaxed when the order follows a person across unknown future facilities. Defenders responded that the court must still find probable cause as to the target and must still find that the target’s conduct may thwart surveillance, and that without the authority, surveillance of sophisticated targets would be trivially easy to evade. The 2015 statute reauthorized the provision without structural change, which reflected a legislative judgment that the existing safeguards, court approval, probable cause as to the target, and the thwarting finding, were sufficient.
The lone wolf provision has a different history. It comes not from the 2001 law but from Section 6001 of the Intelligence Reform and Terrorism Prevention Act of 2004, codified at 50 U.S.C. 1801(b)(1)(C). The provision closed a gap in the original 1978 definition of an “agent of a foreign power.” That definition had assumed that a person engaged in international terrorism would be acting for a foreign power, a government, faction, or terrorist group. But the experience of the early 2000s suggested that individuals could engage in international terrorism without a demonstrable link to any group, inspired by an ideology rather than directed by an organization. The lone wolf provision allowed surveillance of a non-U.S. person engaged in international terrorism even when the government could not prove the person was acting for a known foreign power.
The provision was, by most accounts, rarely used. Its significance was as a backstop: it ensured that the absence of a group affiliation could not, by itself, place a terrorism suspect beyond the reach of the surveillance law. Like the other two powers, it carried a sunset, which is why it appeared in the 2015 debate at all. There was no bulk collection under the lone wolf provision and no controversy comparable to the metadata program; it was bundled into the statute because its sunset date coincided with the others, and Congress chose to address all three expiring powers in a single vehicle.
The bundling had consequences. Because the three powers shared a sunset, the fight over bulk collection determined the fate of all three. The roving and lone wolf provisions, which might otherwise have been reauthorized without controversy, went dark during the roughly one-day lapse and were restored in the same June 2, 2015 vote that ended the bulk program. They then expired again together on March 15, 2020, without reauthorization. The episode illustrates a recurring feature of sunset design: bundling forces joint consideration, which can be efficient, but it also means that controversy over one provision can endanger the others, and that the lapse of one becomes the lapse of all.
An Operational Autopsy of the Replacement Program
The replacement program’s end deserves a closer look, because it is the empirical core of this article’s claim. The call detail records system created by Title I of the June 2, 2015 statute did not end in a blaze of legislative repeal or judicial invalidation. It ended the way complex technical systems often end: with an accumulation of operational problems that made continued operation indefensible, followed by a quiet administrative decision to stop.
The first public sign of trouble came in June 2018, when the National Security Agency announced that it was purging three years of call detail records. The cause was described as technical irregularities: flaws in the production process had caused providers to deliver records beyond what the court’s orders had authorized, and the agency found itself holding data it had no legal right to hold. The critical detail was the agency’s inability to segregate the properly obtained records from the over-collected ones. In a well-functioning system, an over-collection incident would result in the deletion of the excess and the retention of the lawful remainder. Here, the agency concluded that it could not reliably distinguish the two categories, and it therefore deleted the entire three-year set. The purge was both a remedy and an admission: a remedy for the unlawful retention, and an admission that the system’s controls were inadequate to prevent or correct the error.
The irony was sharp. The 2015 statute had been written to end the indiscriminate acquisition of telephone records, replacing a legally dubious bulk program with a precisely authorized targeted one. Yet the replacement had itself acquired records beyond its authorization, not through a sweeping legal interpretation but through technical failure. The distinction between legal overreach and technical malfunction matters for assigning responsibility, but it does not matter much for public confidence. In both cases, the government ended up holding telephone records it should not have held, and in both cases the remedy was deletion. The 2018 purge suggested that the problem with the old program had not been only its legal theory; the machinery of large-scale records acquisition was itself prone to exceeding its bounds.
The end came in early 2019. The agency suspended the call detail records program entirely and deleted the records it had collected under it. The program was not renewed. The decision was administrative, not legislative: the statute’s Title I authority remained on the books, but the collecting agency concluded that the program’s operational difficulties outweighed its intelligence value. Less than four years after its creation, the substantive centerpiece of the 2015 reform had ceased to operate.
Several factors likely contributed to the judgment, though the agency’s public statements emphasized the operational problems. The intelligence value of the program had always been contested; the oversight board’s January 2014 findings about the predecessor program’s limited value cast a long shadow, and there is no public evidence that the replacement proved dramatically more productive. The compliance burden was significant: every production required court approval, every selector required the dual showing, and the provider handoffs that the design depended on proved technically fragile. And the political context had shifted: with the bulk program ended and the replacement struggling, there was little constituency for investing in fixing a system that few defended as essential.
The non-renewal is as significant as the suspension. When the program’s authorization came up again, the agency did not seek to continue it. This was not a case of Congress refusing to extend a power the executive wanted; it was a case of the executive declining to use a power Congress had granted. The distinction matters for the article’s thesis. The replacement was not killed by its opponents. It was abandoned by its operators, which suggests that the negotiated collection compromise at the heart of the 2015 statute was less robust, in practice, than either its supporters or its critics had assumed.
The contrast with the institutional reforms is the point. The amicus panel did not depend on provider handoffs, technical systems, or intelligence productivity judgments. The declassification review duty did not require complex operational machinery. Both were procedural rules about how the court operates, and both continued to function while the collection program was being wound down. The lesson is not that collection programs always fail; it is that procedural reforms and collection programs have different failure modes, and the procedural ones proved more durable.
The Road from Disclosure to Statute
The two years between the June 2013 disclosures and the June 2015 signing can be read as a single narrative, and reading it whole shows how the statute’s final shape was determined less by abstract principle than by the sequence of events that forced each decision. The road had five milestones, and each one narrowed the options available at the next.
The first milestone was the disclosure itself, on June 5, 2013, when The Guardian published the surveillance court order to Verizon Business Network Services. Before that date, the bulk program was a secret known to a small circle; after it, the program was a public controversy. The disclosure did not settle any legal question, but it created the political conditions without which no reform would have occurred. The history of surveillance law suggests that secret programs are rarely reformed while they remain secret, because the constituency for reform cannot organize around what it does not know. The June 2013 order supplied the knowledge.
The second milestone was the oversight board’s report on January 23, 2014. The report translated public outrage into official findings: the program lacked a viable legal foundation, it raised constitutional concerns, and it had shown only limited value. The significance of the date is that it gave reformers an authoritative source more than a year before the sunset, which meant the 2015 debate did not have to relitigate the basic facts. When legislators argued in May 2015 about whether bulk collection worked, they could cite the board’s conclusion that it had never made the concrete difference its defenders claimed. The report did not compel Congress to act, but it deprived the program’s defenders of their strongest arguments.
The third milestone was the legislative stalemate of 2013 and 2014. An earlier bill bearing the same short title was introduced in 2013 and went nowhere. Competing proposals circulated through 2014 without reaching the President’s desk. The stalemate is instructive: even with the disclosures public and the oversight board’s report published, Congress could not assemble a majority for reform in the absence of a deadline. The executive branch defended the program, the intelligence committees were divided, and the ordinary legislative process, with its multiple veto points, favored the status quo. The lesson is that exposure and argument are not always sufficient; sometimes the calendar must do what persuasion cannot.
The fourth milestone was the Second Circuit’s decision on May 7, 2015. The ruling arrived less than a month before the sunset, and its timing was as important as its content. By holding that Section 215 had never authorized bulk collection, the court removed the central premise of the clean-reauthorization position: that Congress had already blessed the program and needed only to extend it. After May 7, a vote for clean reauthorization was a vote to extend a program a federal appeals court had declared unlawful. The ruling did not dictate the replacement’s design, but it foreclosed the simplest path for the program’s defenders and made the reform bill the only viable vehicle for keeping the authorities alive.
The fifth milestone was the Senate’s procedural fight in late May 2015. The May 23 cloture failure, the Paul filibuster, the failed McConnell two-month extension, and the resulting roughly one-day lapse compressed the remaining choices into a single question: the House bill or nothing. The lapse itself became evidence, demonstrating that the authorities could go dark without catastrophe and shifting the burden of proof to the defenders of the old program. On June 2, 2015, the Senate passed H.R. 2048 by 67 to 32, and President Barack Obama signed it that evening.
Read together, the five milestones show a pattern. Each one reduced the set of plausible outcomes: disclosure made the status quo politically untenable, the oversight report made it intellectually untenable, the stalemate showed that only a deadline would force action, the court ruling made clean reauthorization legally untenable, and the Senate fight made the House bill the only alternative to expiration. The statute that emerged was not the product of a single moment of consensus but the survivor of a two-year process of elimination.
Sunsets as a Legislative Technology
The 2015 statute cannot be understood without understanding sunsets, the legislative technology that made the entire episode possible. A sunset is an expiration date written into a law: the provision ceases to have legal effect on the specified date unless Congress affirmatively extends it. Sunsets reverse the ordinary presumption of legislation, which is permanence. Instead of requiring opponents to assemble a majority for repeal, a sunset requires supporters to assemble a majority for extension. The difference is decisive in a system with as many veto points as the American legislature.
Congress built sunsets into the surveillance powers for a specific reason: the powers were extraordinary, enacted in the aftermath of the September 2001 attacks, and their authors wanted to ensure periodic reconsideration. The business records provision, the roving wiretap authority, and eventually the lone wolf provision all carried expiration dates. Each extension was an opportunity for Congress to revisit the bargain, and each extension passed with the understanding that the question would return. The sunset did not guarantee reform, but it guaranteed revisitation, which is the precondition for reform.
The 2015 episode shows both the strength and the weakness of the device. The strength is that the May 31, 2015 sunset forced action when two years of debate had not. Without the deadline, the bulk program might have continued indefinitely under its secret interpretation, sustained by legislative inertia. The weakness is that the deadline also created the conditions for the lapse: when the Senate’s procedural fight consumed the available time, the powers expired not because a majority wanted them to expire but because no majority could act in time. The roughly one-day gap was the price of the forcing mechanism.
Grandfather clauses are the technology that makes sunsets tolerable. When Section 215 expired at 11:59 p.m. on May 31, 2015, it reverted to its pre-2001 form, but the grandfather provisions preserved the expiring authorities for investigations already underway. The design reflects a judgment about fairness and practicality: it would be disruptive and arguably unjust to terminate ongoing investigations midstream because Congress missed a deadline, but there is no comparable reason to allow new investigations to begin under expired powers. The clauses drew a line between the past and the future, preserving the former while foreclosing the latter.
The 2020 sunset repeated the pattern on a longer timeline. The 2015 statute extended the three powers with a new expiration date, March 15, 2020, and when that date arrived without reauthorization, the powers lapsed again. This time there was no replacement vehicle and no dramatic restoration; the authorities simply expired. The difference between 2015 and 2020 illustrates how sunsets interact with political context. In 2015, the sunset forced a choice between reform and expiration, and Congress chose reform. In 2020, the sunset forced the same choice, and Congress, through inaction, chose expiration. The technology does not determine the outcome; it determines that there will be an outcome, by a date certain.
There is a broader lesson for the series. The amendment stage of a statute’s life is often triggered by a sunset, because sunsets are the moments when the legislature must confront what the law has become. The USA FREEDOM Act of 2015 is the clearest example in the series of a sunset functioning as designed: it forced Congress to examine a secret interpretation, to weigh the evidence of effectiveness, and to write a new framework under pressure. The fact that the framework’s collection program was later abandoned does not diminish the sunset’s role; if anything, it confirms that the 2015 reexamination was substantive rather than ceremonial. Congress looked, decided, and built, and what it built was then tested by reality.
The Providers’ Side of the Design
The 2015 statute is usually described from the government’s perspective: what the FBI must show, what the court must approve, what the agency may obtain. But the replacement design also transformed the role of the telephone providers, moving them from the periphery of the system to its center. Understanding the providers’ side shows why the new architecture was both more privacy-protective and more operationally fragile than the program it replaced.
Under the bulk program, the providers’ role was simple and passive. The surveillance court’s bulk orders required carriers to deliver telephone metadata to the government on an ongoing, daily basis, and the carriers complied. Once the records were delivered, the government’s database took over: storage, organization, and querying were all government functions. The providers were the source of the records but not the custodians of the system. Their operational burden was limited to the mechanics of daily production.
The 2015 statute inverted this arrangement. Under Title I, providers retain their own call detail records in the ordinary course of business, and the government obtains specified records only when the surveillance court approves a specific selection term. There is no government database. The providers are the custodians of the records, and the government is a requester, entitled to receive only what a court order specifies. The shift was deliberate: the statute’s drafters concluded that the existence of a government-held database of everyone’s calls was itself the danger, regardless of the rules governing its use, and the only way to eliminate the danger was to ensure that no such database existed.
Retention “in the ordinary course of business” is doing quiet work in the design. The statute does not impose a general data-retention mandate on providers; it does not require them to keep records they would otherwise discard. Instead, it builds on the records they already maintain for billing, network management, and other business purposes. The government’s access is thus bounded by what the providers happen to hold, which introduces a natural limit on the historical reach of any production. Records the provider has discarded in the ordinary course are simply unavailable, and the statute does not require their resurrection.
When the court approves a specific selection term, the provider’s obligations activate. The provider must isolate the call detail records tied to the selector and produce them within the deadlines the court’s order specifies. This requires technical systems capable of querying the provider’s records by the identifiers the statute allows, person, account, address, personal device, or other specific identifier, and of delivering the results in a form the government can use. The systems must also respect the order’s limits: production is confined to the selector and its two-hop neighborhood, bounded by the 180-day window, and subject to the court’s minimization requirements as implemented by the government on receipt.
The definition of the specific selection term protects providers as well as subscribers. Under 50 U.S.C. 1861(k)(4), the term’s use must be limited to the greatest extent reasonably practicable, and the definition excludes providers that are not subject to the order. A provider that does not hold the selector’s records cannot be swept into production on the theory that it might hold something useful. The exclusion of broad geographic regions serves a similar function: an order cannot require every provider in an area code or metropolitan region to produce, which would recreate bulk collection through the multiplication of targeted orders. Each provider’s obligation is tied to the selector, not to geography.
The infrastructure burden of the new system sits with industry. Storage, retrieval, and compliance, the functions the government performed for itself under the bulk program, became provider functions performed under court order in the replacement design. The statute does not compensate the providers for these costs as a general matter; the obligations are the price of operating a communications network under the foreign intelligence law. The trade was explicit in the legislative design: providers bear the operational burden, and in exchange the public gets a system in which no government database of everyone’s calls exists to be queried, misused, or breached.
The 2018 over-collection incident revealed the fragility of this arrangement. The technical irregularities that caused providers to deliver records beyond what the court’s orders permitted were failures at the handoff between provider systems and government receipt, the exact point where the new design concentrated its complexity. Under the bulk program, the handoff was simple: deliver everything, daily. Under the replacement, the handoff required precise isolation of selector-tied records within defined hop limits and time windows, executed through systems that had to interpret court orders correctly. The 2018 purge showed that this precision could not always be achieved, and the agency’s inability to segregate the over-collected records from the lawful ones showed that the failure modes were not gracefully contained.
The providers’ relationship to the system is compelled, not voluntary. The statute authorizes the court to order production, and providers that receive such orders must comply, subject to the usual mechanisms for challenging unlawful orders. This is not a partnership or a data-sharing agreement; it is a legal duty backed by the court’s authority. The distinction matters because it locates accountability: when over-collection occurs, the question is whether the provider’s systems correctly implemented the court’s order, and the remedy runs through the court that issued it.
When the program was suspended in early 2019, the providers’ role ended as quietly as it had operated. There were no more court orders to implement, no more selector-tied productions to deliver. Retention returned to its ordinary business purposes, governed by the providers’ own practices and the generally applicable law rather than by the foreign intelligence statute’s production regime. The quiet ending is characteristic of the replacement as a whole: a system built on provider-held records, when abandoned, leaves nothing behind, because there was never a government database to dismantle.
The Title as an Argument
The statute’s full short title is itself worth reading: the Uniting and Strengthening America by Fulfilling Rights and Ensuring Effective Discipline Over Monitoring Act of 2015. The capital letters spell USA FREEDOM, a backronym in the tradition of the 2001 antiterrorism law whose patriotic abbreviation it echoes. Congressional drafters choose such titles deliberately, and this one repays attention, because it states the coalition’s theory of the bill in the bill’s name.
The operative words are “Fulfilling Rights” and “Ensuring Effective Discipline Over Monitoring.” The title pairs two ideas that are often treated as opposites: the protection of rights and the effectiveness of intelligence. The claim embedded in the title is that these are not opposites at all, that discipline over monitoring, meaning judicial supervision, particularized suspicion, and the end of indiscriminate collection, makes surveillance more legitimate and therefore more sustainable. The title argues, before the first section is even read, that reform is not the enemy of security but its precondition.
The contrast with the 2001 law’s title is instructive. The earlier title emphasized the provision of tools against terrorism, an expansionist framing suited to the aftermath of the September 2001 attacks. The 2015 title emphasizes rights and discipline, a constraining framing suited to the aftermath of the 2013 disclosures. The shift in titles mirrors the shift in the statute’s life stage: from the grant of powers to the amendment of powers, from the question of what the government may do to the question of how what it does is supervised. The series thesis, that statutes have life stages, is visible even in the naming conventions.
Titles do not change operative provisions, and no court has ever decided a case on the basis of a short title’s poetry. But titles signal the self-understanding of the coalition that passes them, and this coalition understood itself as strengthening the surveillance framework by disciplining it. Whether the discipline worked is the question the rest of this article has examined. The title promised that rights and effectiveness could be fulfilled together; the history, from the 2018 purge to the 2019 suspension to the 2020 expiration, suggests that the promise held better for the institutional reforms than for the collection program. The title remains, in any case, as the most concise statement of what the 2015 Congress thought it was doing.
The before-during-after table
| Phase | Authority relied on | The court’s role | Status |
|---|---|---|---|
| The bulk program, in operation when disclosed on June 5, 2013 | Section 215 of the foreign intelligence statute, 50 U.S.C. 1861, read to permit bulk orders | The surveillance court approved bulk production orders on a recurring basis | Ended by the June 2, 2015 statute, which barred bulk collection under this authority |
| The lapse period, May 31 to June 2, 2015 | Section 215 in its pre-2001 form, with grandfather clauses covering open investigations | No new bulk orders were available; the court acted only within the narrowed, grandfathered powers | A roughly one-day lapse following the May 23, 2015 cloture failure; restored at signing on June 2, 2015 |
| The replacement program, June 2015 to early 2019 | Title I of the 2015 statute: provider-held call detail records obtained on court approval of a specific selection term | The court approved each selector and each production; two hops permitted | Suspended by the collecting agency in early 2019 after the June 2018 purge; the records were deleted; not renewed |
| The aftermath, March 15, 2020 | None; the business records authority expired on March 15, 2020, without reauthorization | No live business records authority for the court to administer | Expired; the roving wiretap and lone wolf powers expired the same day |
Frequently Asked Questions
Q: What did the USA FREEDOM Act change?
The statute, signed on June 2, 2015, ended the government’s bulk collection of domestic telephone metadata under the business records provision and replaced it with a narrower system. Under the replacement, telephone providers keep their own records, and the government may obtain specified call detail records only after winning approval from the surveillance court for a specific selection term tied to a suspected foreign power or its agent. The measure also barred the broad collection the 2013 disclosures had revealed, added minimization rules requiring prompt destruction of irrelevant information, and created two lasting institutional changes: a panel of outside advisers the surveillance court must appoint in significant cases, and a duty for intelligence officials to review significant court opinions for declassification and public release. It left other surveillance powers, including the separate Section 702 program, untouched.
Q: Did the USA FREEDOM Act end bulk collection?
It ended one specific bulk program: the government’s bulk collection of domestic telephone metadata under Section 215, the business records provision. It did not end every form of bulk collection. The separate Section 702 program, authorized under 50 U.S.C. 1881a, and collection conducted overseas under Executive Order 12333 were not amended or restricted by the statute, apart from a narrow limit on the use of information derived from unlawfully certified 702 procedures. The honest answer is therefore narrower than either side’s slogan: the government could no longer hold a bulk domestic telephone metadata database under Section 215, but other collection authorities continued under their own separate legal rules.
Q: Why did surveillance authorities lapse before the USA FREEDOM Act?
Three provisions, Section 215, the roving wiretap power in Section 206, and the lone wolf provision from Section 6001 of the 2004 intelligence reform law, were all scheduled to expire at 11:59 p.m. on May 31, 2015. On May 23, 2015, the Senate failed to invoke cloture on the House-passed reform bill, falling to 57 votes against the 60 required, after a filibuster by Senator Rand Paul. A separate two-month clean reauthorization pushed by Senator McConnell then failed 45 to 54. With no extension in place, the authorities expired at midnight and stayed dark for roughly one day. On June 2, 2015, the Senate passed the reform bill 67 to 32, and President Barack Obama signed it that evening, restoring the powers in modified form.
Q: What is a specific selection term in the USA FREEDOM Act?
It is the selector the government must present to the surveillance court before obtaining any call detail records, and the statute defines it narrowly. Under 50 U.S.C. 1861(k)(4), added by Section 108 of the measure, a specific selection term must specifically identify a person, account, address, or personal device, or another specific identifier, and its use must be limited to the greatest extent reasonably practicable. The definition excludes broad geographic regions and providers that are not subject to the order. To win approval, the FBI must also show reasonable grounds that the records sought are relevant to an investigation and a reasonable, articulable suspicion that the term is associated with a foreign power or its agent engaged in international terrorism. This selector requirement is what replaced bulk collection.
Q: What is the FISA court amicus panel from the USA FREEDOM Act?
Section 401 of the statute created a standing pool of outside advisers for the surveillance court, codified at 50 U.S.C. 1803(i). The presiding judges of the trial-level surveillance court and its review court jointly designate at least five individuals eligible to serve as amicus curiae, chosen for expertise in privacy and civil liberties, intelligence collection, or communications technology, who must be eligible for access to classified information. The court must appoint one of them in any application or review presenting a novel or significant interpretation of the law, unless it finds the appointment inappropriate. The idea was to give the court an adversarial voice arguing against the government in the cases that matter most, changing an institution that had previously heard only one side.
Q: Did Edward Snowden cause the USA FREEDOM Act?
The disclosures associated with Edward Snowden, which began on June 5, 2013 with the publication of a surveillance court order to Verizon Business Network Services, exposed the bulk telephone metadata program and made reform politically unavoidable. Without that exposure, the secret interpretation of the business records provision would likely have continued unchallenged. But Congress wrote and passed the statute, the courts shaped it through the May 2015 appellate ruling, and the executive branch negotiated its final form. The fair description is that the disclosures were the trigger and the indispensable cause of the public debate, while the act itself was the product of the legislative, judicial, and executive branches responding to what the disclosures revealed.
Q: Is the USA FREEDOM Act still in force?
Only in part. The collection program the statute created was suspended by the collecting agency in early 2019, its records were deleted, and the program was never renewed. Then, on March 15, 2020, the underlying business records authority itself expired without reauthorization, taking the roving wiretap and lone wolf provisions with it. Those expirations were not repeals of the 2015 statute; they were sunsets of the powers it had reauthorized. The institutional reforms stand on different footing: the outside adviser panel and the declassification review duty were written into permanent sections of the foreign intelligence statute, not into the provisions that expired in 2020.
Q: What happened to the call detail records program under the USA FREEDOM Act?
Title I of the statute created it in June 2015 as the replacement for government-held bulk collection: providers kept the records, and the government obtained specified records only with court approval of a specific selection term, limited to two hops of contacts. In June 2018, the collecting agency purged three years of records after discovering technical irregularities that had caused over-collection beyond what the court had authorized. In early 2019, the agency suspended the program entirely and deleted the call detail records it held, and the program was not renewed when its authorization came up again. The episode is the reason this article’s central claim holds: the procedural reforms outlived the collection program they accompanied.
Q: How does the FISA court amicus panel operate in practice?
The presiding judges of the two surveillance courts jointly designate at least five eligible advisers, and the statute directs the court to appoint one whenever an application or review presents a novel or significant interpretation of the law, unless the court finds appointment inappropriate in that case. Once appointed, the adviser receives access to the classified materials, files briefs arguing against the government’s position, and may participate in proceedings as the court permits. Designees must hold expertise in privacy and civil liberties, intelligence collection, or communications technology, and must be eligible for classified access. The panel does not give the adviser party status or a veto; it gives the judges an adversarial perspective in an institution that previously heard only the government’s side.
Q: What does the declassification review requirement cover?
Section 402 of the statute added Section 602 to the foreign intelligence law, codified at 50 U.S.C. 1872. It directs the Director of National Intelligence, consulting with the Attorney General, to conduct a declassification review of every surveillance court and review court decision, order, or opinion that includes a significant construction or interpretation of any provision of law, and to make each one publicly available to the greatest extent practicable. Redacted releases are permitted where national security requires, and the duty may be waived for a particular opinion only with an unclassified summary explaining why. The requirement was aimed at the body of secret law that had grown up around the bulk program, where major legal interpretations bound the government but were invisible to the public.
Q: What limit did the act place on the use of information collected under Section 702?
Almost none, by design. The statute did not amend, restrict, or reauthorize Section 702, the separate authority under 50 U.S.C. 1881a for targeting non-U.S. persons abroad, and it left untouched the practice of querying communications collected under that authority. Its single Section 702-related provision, Section 301, limited the government’s use of information derived from targeting or minimization procedures that had been unlawfully certified, a narrow evidentiary rule rather than a programmatic reform. Readers sometimes conflate the two authorities because both involve the surveillance court, but the 2015 measure’s reforms applied to the business records, roving, and lone wolf powers, while Section 702 continued under its own separate legal framework.
Q: How do the two hops work in the call detail records program?
The two-hop rule defines how far the government may travel through telephone records from an approved starting point. The first hop covers the call detail records directly associated with the court-approved specific selection term, meaning the numbers in contact with the suspected identifier. The second hop covers the records of the persons identified in the first hop, meaning the contacts of those contacts. No third hop is permitted, and each production requires the same court approval, the same relevance showing, and the same reasonable, articulable suspicion. Minimization rules then require prompt destruction of any records that do not contain foreign intelligence information. The hop limit is what keeps a targeted query from expanding back into bulk collection.
Q: What grandfather clauses applied when the authorities lapsed in 2015?
When Section 215 expired at 11:59 p.m. on May 31, 2015, it reverted to its pre-2001 form, but grandfather clauses preserved the expiring powers for investigations that were already open. Any investigation begun before the sunset could continue under the lapsed authorities as if they had not expired, so ongoing counterterrorism and counterintelligence work did not have to stop midstream. What the lapse foreclosed was new reliance on the expired powers: no new bulk orders, no new roving wiretap applications, and no new lone wolf applications could be initiated during the roughly one-day gap. The clauses are the reason the lapse was a controlled, brief interruption rather than a collapse of active cases.
Q: Why did the NSA purge three years of call detail records in 2018?
In June 2018, the National Security Agency announced that it was deleting three years of call detail records it had obtained under the replacement program because of technical irregularities. The agency had received records it was not authorized to receive: due to technical problems in the production process, providers had delivered data beyond what the court’s orders permitted, and the agency could not reliably separate the properly obtained records from the over-collected ones. Rather than retain a database of uncertain legality, it purged the entire three-year set. The episode damaged confidence in the program’s administration and foreshadowed its suspension the following year.
Q: When did the NSA delete the call detail records it had collected?
In early 2019, the National Security Agency suspended the call detail records program and deleted the records it had collected under it. The suspension followed the June 2018 purge of three years of over-collected records and reflected the agency’s judgment that the program’s operational difficulties outweighed its intelligence value. The program was then not renewed, so the provider-held, court-approved system created by Title I of the 2015 statute ceased to operate less than four years after it began. The deletion and non-renewal were administrative decisions by the collecting agency, not a statutory repeal, which is why the underlying legal authority could still have supported a future program until it expired in March 2020.
Q: What minimization rules apply to call detail records?
The statute requires the government to destroy promptly any call detail records it obtains that do not contain foreign intelligence information, so irrelevant records may not be stockpiled for future use. The definition of call detail records itself does the first round of minimization: it excludes the contents of communications, the subscriber’s name, address, and financial information, and cell-site and GPS location data. What remains is transactional metadata, numbers, identifiers, and the time and duration of calls, and even that may be kept only under court-approved procedures designed to limit retention and dissemination. These rules were part of the answer to the old program, where the government held everything indefinitely.
Q: What is the lone wolf provision that lapsed in 2015?
The lone wolf provision, Section 6001 of the Intelligence Reform and Terrorism Prevention Act of 2004, codified at 50 U.S.C. 1801(b)(1)(C), allowed surveillance of a non-U.S. person engaged in international terrorism even when the government could not prove the person was acting for a known terrorist group or foreign power. It closed a gap in the original 1978 definition of an agent of a foreign power, which had assumed a link to a group or government. Like Sections 215 and 206, it carried a sunset, so it expired at 11:59 p.m. on May 31, 2015, was restored by the June 2, 2015 statute, and then expired again on March 15, 2020, without reauthorization.
Q: What is the roving wiretap authority and why did it lapse?
Section 206 of the 2001 antiterrorism law, the roving wiretap provision codified at 50 U.S.C. 1805(c)(2)(B), lets a surveillance order follow the target rather than a specific telephone line or facility, so investigators do not need a new court order each time a suspect switches phones. It lapsed for the same reason the other two powers did: Congress had given all three provisions a common sunset date, and the Senate’s procedural fight in late May 2015 left no extension in place when 11:59 p.m. on May 31 arrived. The June 2, 2015 statute restored it, and it expired again alongside Section 215 on March 15, 2020, without reauthorization.
Q: Who sponsored the USA FREEDOM Act in Congress?
Representative Sensenbrenner introduced the House vehicle, H.R. 2048, on April 28, 2015, and the House passed it on May 13, 2015, by 338 to 88. Senators Lee and Leahy introduced a Senate companion, S. 1123, but that bill was never voted on and did not become law. After the Senate’s late-May procedural fight allowed the authorities to lapse, the Senate took up the House bill directly, rejected all three amendments offered by Senator McConnell, and passed H.R. 2048 on June 2, 2015, by 67 to 32. President Barack Obama signed it that evening as Public Law 114-23.
Q: What costs and duties did the act place on telephone providers?
The statute moved the burden of holding telephone records from the government to the providers. Companies must retain their own call detail records in the ordinary course of business and must be able to produce specified records promptly when the surveillance court approves a specific selection term, which requires systems capable of isolating the records tied to a selector and delivering them within the court’s deadlines. The government no longer maintains its own bulk database, so the infrastructure cost of storage, retrieval, and compliance sits with industry. The trade was deliberate: providers bear the operational burden, and in exchange the public gets a system in which no government database of everyone’s calls exists to be queried.
The Amendment Stage and Its Natural Experiment
Statutes have life stages, and the amendment stage is the one in which a law is tested against the consequences of its own interpretation. The USA FREEDOM Act of 2015 is the amendment stage of the business records provision, and it arrived with an unusually clean natural experiment attached. Because the underlying authorities actually lapsed, the same agencies, the same investigators, and the same threats were observed under three conditions in sequence: the bulk program, the roughly one-day gap with no powers at all, and the reformed replacement. Legislators rarely get to see what the absence of an authority looks like, and they almost never get to compare the before, the during, and the after. The lapse made the comparison unavoidable, and the comparison favored reform: the gap proved survivable, the replacement proved workable, and the old program’s defenders lost the argument that only bulk collection could do the job.
The reform that outlived its program: the durable achievements of the 2015 act are institutional, the outside advisers and the declassification requirement, while the substantive program it created was abandoned within a few years, which suggests that procedural reforms to a secret court survive better than negotiated collection compromises.
That claim is worth stating plainly because it runs against the usual way reform is judged. Reform is typically measured by the programs it creates, and by that measure the 2015 statute’s record is mixed: it built a replacement system that lasted less than four years before suspension and deletion. But the statute also changed the institution that approves surveillance, giving its judges an adversarial voice and its legal interpretations a path into public view, and those changes did not depend on any program’s operational success. Procedural reforms survive because they alter how decisions get made rather than what gets collected, and how decisions get made is harder to abandon than a database that keeps malfunctioning.
The natural experiment point deserves a final emphasis, because it is what makes this episode unusual in the series. Most amendments to surveillance law are negotiated under the shadow of the existing program: the powers continue while Congress debates, and the debate is shaped by warnings about what might happen if they do not. Here, the powers actually stopped, if only for roughly one day, and the warnings were tested against reality. The grandfather clauses kept open investigations running, no new authorities were available, and the government continued to function. That lived experience could not prove that the reformed framework would work forever, but it shifted the burden of proof. After the lapse, it was the defenders of bulk collection who had to explain why the old program was necessary, not the reformers who had to explain why the country would survive without it.
The comparison also clarifies what the 2015 compromise cost and what it bought. The cost was the loss of the bulk database as an investigative resource, a capability whose value the oversight board had already found to be limited and whose legal foundation the appeals court had already rejected. The purchase was a system in which every acquisition of telephone records required individualized judicial approval tied to a terrorism-related suspicion, plus two permanent changes to the secret court’s procedures. Measured against the alternative of clean reauthorization, which would have preserved the bulk program the court had just declared unlawful, the compromise bought legality and oversight at the price of a capability of doubtful worth. Measured against the alternative of abolition, which would have ended the business records authority entirely, it preserved a targeted tool for investigators willing to meet the court’s standards.
There is a final reason the institutional reforms deserve the emphasis this article gives them. Collection programs are creatures of technology and threat perception, and both change. The bulk program was a response to the communications landscape of the early 2000s; the replacement was a response to the political landscape of 2015; neither was built to last decades. Procedural reforms operate at a different level of abstraction. An adversarial voice in significant cases and a duty to publish significant legal interpretations are valuable regardless of what is being collected, because they address the standing risks of secret adjudication rather than the transient details of any one program. The 2015 statute’s drafters may not have foreseen the replacement’s abandonment, but they built the parts of the law that mattered most to survive it.
The expiration also clarified the relationship between the 2015 statute and the broader surveillance landscape. For five years, the business records authority had operated in its reformed, post-2015 form, supporting the provider-held call detail records system for the roughly four years that system functioned. When the authority lapsed in March 2020, it did not merely end a dormant power; it closed the legal chapter that had begun with the 2001 expansion, continued through the secret bulk program, and ended with the targeted replacement and its abandonment. The arc from expansion to secrecy to exposure to reform to expiration is complete, and it is one of the few complete arcs in the series: a power granted, used beyond its text, corrected by courts and legislators, and finally allowed to expire.
Readers who want to trace each provision of the statute to its source can work through the measure methodically in the VaultBook legislation study notebook, which is built for exactly this kind of section-by-section study. The statute rewards close reading: its most important moves are definitions, like the specific selection term, and procedures, like the amicus appointment duty, rather than sweeping declarations. That, in the end, is the character of the amendment stage. It does not repudiate the original law or vindicate it. It revises the machinery, watches what happens, and leaves behind the parts that worked. In this case, the parts that worked were the ones that opened the secret court, however slightly, to argument and to light. The bulk program is history; the procedures endure and continue to operate.