How Much Surveillance Can a Democracy Tolerate? — Security, Privacy and Emergency Powers
When does security surveillance cross democratic limits? A review of US, British, and European practices, judicial oversight, and later reforms.
AI-generated symbolic editorial illustration, not a documentary image.
Democracies like to describe privacy as a fundamental right. At the same time, some of them possess the most powerful intelligence systems in history. That is not, by itself, a contradiction.
A state has a legitimate responsibility to prevent terrorist attacks, espionage, sabotage, cyberattacks and other serious threats. No serious privacy framework requires security services to operate blindfolded. The real question is different.
How much surveillance can a democracy normalise before it begins treating its own exception as something it would not accept from others?
This is where Western rhetoric about privacy becomes much less simple.
Democratic surveillance is not the same as authoritarian surveillance
A clear boundary has to be drawn first. Britain’s GCHQ, the U.S. National Security Agency or European security services are not automatically equivalent to political-surveillance apparatuses in systems without independent courts, free media, parliamentary opposition or effective legal remedies. That difference is real.
Western democracies have courts, parliamentary committees, independent oversight bodies, public reports, judicial proceedings, declassification processes and at least some possibility that a programme can be changed or terminated after exposure. That is precisely why the cases in this article matter.
If the system itself finds that some of its methods crossed acceptable boundaries, criticism cannot be dismissed merely by invoking “national security”. Surveillance in a democracy is not a problem simply because it exists.
The problem begins when exceptional powers become ordinary, legal exceptions become permanent, and effective oversight arrives only after an intrusion has already been embedded in the system for years.
Section 215: when “relevant” came to mean almost all telephone records
One of the clearest examples is the U.S. bulk telephone-metadata programme conducted under Section 215 of the USA PATRIOT Act.
The programme did not collect the content of telephone calls. It collected call metadata. That is an important distinction. Metadata are not the same as the content of a communication.
But large networks of telephone numbers, times and call durations can reveal a detailed pattern of a person’s relationships and activities. The privacy question therefore did not disappear simply because the government was not listening to the content of every call.
The strongest institutional criticism of the programme came from the U.S. Privacy and Civil Liberties Oversight Board.
In 2014, a majority of the Board concluded that the programme lacked a viable legal foundation under Section 215, raised constitutional concerns, posed serious risks to privacy and civil liberties, and had shown only limited value. The Board was unaware of any instance in which the programme directly discovered a terrorist plot, or in which it directly contributed to disrupting one. It identified one case in which the data may have helped identify a previously unknown suspect, but that person was not planning an attack and the Board believed the FBI might have identified him without those records. [1]
An important corrective: the recommendation to end the programme was not unanimous. Two members of the Board did not join it. That does not change the findings about the programme’s limited demonstrated value.
But it does show that a serious dispute existed even within the oversight body over how much preventive value should be assigned to a large database that might become critical only in a future crisis. That is where the proportionality test becomes difficult.
If a state collects data on an enormous scale, the standard for justification should be higher than the claim that the data might someday prove useful.
The correction came only after years of disputed practice
The USA FREEDOM Act of 2015 prohibited the previous form of bulk collection under Section 215 and created a different mechanism under which the data remain with providers and the government obtains them through more targeted, court-approved requests. The earlier bulk-collection model ended after a transition period in November 2015. [2]
The replacement system was not trouble-free either. In 2018, the NSA publicly acknowledged technical irregularities that had caused it to receive records it was not authorised to obtain. Because properly acquired records could not be reliably separated from improperly acquired ones, the agency decided to delete all call detail records then held under the new system. [3]
In August 2019, the NSA permanently decommissioned the CDR programme. At a congressional hearing, officials explained that the decision followed a weighing of intelligence value, cost, compliance problems and data-integrity issues. [4]
The fact that the programme was later restricted and then abandoned does not lessen the problem of its original scale. On the contrary.
The system required years of secret operation, public controversy, institutional review, legislative reform and new technical failures before the mass-collection model was brought to an end. The correction came after the intrusion, not before it.
Section 702: a different programme, a different problem
Section 702 must not be confused with Section 215. It is not the same programme and, under the U.S. government’s own description, it is not bulk collection in the sense of indiscriminately targeting an entire population.
Section 702 authorises targeted collection against non-U.S. persons reasonably believed to be outside the United States and expected to possess, receive or communicate a defined category of foreign-intelligence information. U.S. persons and people located inside the United States may not be directly targeted under Section 702; so-called reverse targeting — using a foreign target as a pretext to surveil a known American — is also prohibited. [5]
The security argument is strong. Foreign intelligence officers, terrorist organisations, cyberattackers and other foreign actors use American communications infrastructure. Requiring a traditional individual warrant for every foreign target abroad would fundamentally change the way foreign-intelligence collection operates. But a different problem arises. Foreign targets communicate with Americans.
Communications involving U.S. persons can therefore be lawfully acquired incidentally. Under prescribed conditions, the FBI can later search previously collected data using U.S.-person identifiers.
It is this secondary use that lies at the centre of the political dispute critics describe as a backdoor search.
Millions of queries are not theoretical — but the number must be read correctly
In its 2023 report, the PCLOB documented that the FBI reported 3,394,053 queries associated with U.S. persons during the twelve-month period ending 30 November 2021. But the raw number can mislead.
The methodology at the time counted repeated queries. After later deduplication, the figure amounted to 2,964,643 unique query terms. Roughly 1.9 million of the queries were linked to a single major cyber threat and potential victims. [6]
This means the 3.4 million figure does not represent 3.4 million people deliberately spied on by the FBI. Nor does it represent 3.4 million separate suspicions of crime.
Even after that correction, however, a serious question remains: under what conditions should a domestic security agency be allowed to search for its own citizens’ communications inside a database created by a foreign-intelligence programme?
U.S. institutions recognised the problem and introduced technical, procedural and statutory restrictions. The effect is measurable.
For the period from December 2023 to November 2024, ODNI reported 5,518 unique FBI query terms associated with U.S. persons, down from 57,094 in the previous year. In the following period, December 2024 to November 2025, the number rose slightly to 7,413, but remained dramatically below earlier levels. [7] [8]
In a publicly released opinion on certifications from March 2025, the FISA Court also assessed that reforms appeared to be improving the FBI’s compliance with query standards, while still addressing instances of improper designation and the need for further improvement. [9]
This is an important corrective. The surveillance system did not remain unchanged. But the fact that such substantial reforms were necessary, and that the numbers fell by orders of magnitude afterward, confirms that the earlier problem was not merely rhetorical.
A reform that restricts and expands at the same time
The Reforming Intelligence and Securing America Act of 2024 introduced additional restrictions. In its 2026 updated report, the PCLOB describes tighter requirements for certain FBI queries, more training and reporting, new disciplinary rules for non-compliance, and a more mandatory role for amici curiae in Section 702 certifications.
But the same law also expanded certain parts of the system: the definition of some electronic communications service providers, the category of foreign-intelligence interests involving the international production and distribution of certain drugs, and the use of Section 702 information for vetting non-U.S. persons. [10]
That is a recurring problem in surveillance policy. A reform can strengthen safeguards and expand scope at the same time.
So the word “reform” alone does not tell us whether the state became less or more capable of intruding on privacy. The individual provisions matter.
In 2026, the dispute did not end — and a legal paradox emerged
Section 702 acquired an unusual status in 2026. Congress did not enact a permanent reauthorisation before the deadline. After several temporary extensions, the statutory authority expired on 12 June 2026.
Reuters reported before the deadline that existing, already approved certifications could allow a substantial part of collection to continue, even though new requests and provider cooperation could face limitations. [11]
That is what happened. Existing one-year FISA Court certifications approved before the statutory lapse allow surveillance to continue roughly until March 2027. As late as September 2026, debate therefore continued over whether and under what conditions Congress should reauthorise Section 702. [12] [13]
The legal distinction matters:
the statute expired, but the surveillance did not switch off on the same day.
The dispute is therefore not merely historical. It remains an active argument over whether Congress should restore the authority, what safeguards should accompany it, and whether queries seeking U.S.-person communications should require a judicial warrant. This is not a dispute between “security” and “anarchy”.
It is a dispute over where foreign-intelligence collection ends and the domestic investigation of a citizen begins.
Britain: bulk interception is not automatically prohibited
The British case matters because the European Court of Human Rights did not rule that bulk interception is always incompatible with a democratic system.
In Big Brother Watch and Others v. the United Kingdom, the Grand Chamber in 2021 emphasised that bulk interception requires end-to-end safeguards.
The Court required assessment of necessity and proportionality at every stage, independent authorisation at the outset, supervision of implementation and effective independent ex post review.
The old British system under the Regulation of Investigatory Powers Act 2000 did not meet those requirements. The Court found violations of Article 8 of the Convention in the bulk-interception regime and in the acquisition of communications data, as well as violations of Article 10 because of insufficient protections for journalistic material and sources. [14]
This is a very important boundary. The Court did not say a democracy must abandon every form of bulk interception. It said that such an intrusive power requires safeguards that operate from beginning to end.
The British legal framework was later replaced by the Investigatory Powers Act 2016. The Act regulates bulk warrants and uses a so-called double lock: a ministerial decision must also be approved by a Judicial Commissioner before a warrant is issued. The system contains additional rules for examining, retaining and using material, as well as special safeguards for journalistic sources. [15] [16]
In 2024, the Act was amended further in relation to journalistic safeguards, and supervision of the execution of the Big Brother Watch judgment within the Council of Europe system was subsequently closed. [17]
This is a strong example of self-correction. But the legal message remains revealing:
democracy did not abolish bulk interception. It retained it under stricter conditions.
European law: general retention is prohibited in principle — but not without exception
A similar duality appears in European Union law. In 2020, in Privacy International and La Quadrature du Net, the Court of Justice of the European Union held that EU law generally precludes the general and indiscriminate transmission or retention of traffic and location data.
But where a state faces a serious threat to national security that is genuine and present or foreseeable, the Court recognised that general and indiscriminate retention may be ordered for a period limited to what is strictly necessary; the measure may be extended if the threat persists. [18]
That is more precise than the slogan that the EU “bans mass surveillance”. The European standard does not eliminate the security exception. It attempts to confine it legally. The problem therefore moves to different questions: who determines that a threat is serious enough, how long an exceptional measure may remain in force, what independent control exists, and whether the exception really contracts once the concrete threat has passed.
Pegasus: spyware was not only a problem of “others”
Uradni dokument — www.intelligence.senate.gov
Western states speak very strongly about digital surveillance of dissidents, journalists and opposition figures in authoritarian countries. That criticism is often justified.
But the European Parliament’s inquiry into Pegasus and comparable spyware showed that the problem is serious inside the EU as well.
After a year-long investigation, the PEGA committee described abuses of spyware to monitor, intimidate or discredit political opponents, journalists and civil society. The European Parliament subsequently demanded that such tools be used only in exceptional cases, for a limited period and under strict conditions. [19] [20]
The final PEGA report stated that, according to information from NSO Group, Pegasus had been sold in at least 14 EU member states. At the same time, the report distinguished between the sale of the tool and its unlawful use, and identified particular countries in which the committee found abuses or suspected abuses. [21]
This does not mean all fourteen states used Pegasus unlawfully. Nor does it mean that every use of advanced spyware is unjustified.
Targeted surveillance of a person suspected of a serious crime, under independent legal supervision, is fundamentally different from secretly infecting the telephone of a journalist or opposition politician because of their work. But the technology is the same.
What creates the difference is purpose, legal basis, proportionality, independent authorisation, oversight and the availability of a remedy.
Later restrictions do not erase Western spyware abuses
After abuses were exposed, the United States also introduced restrictions against parts of the commercial-spyware industry.
In 2021, the U.S. Department of Commerce added NSO Group and Candiru to the Entity List on the basis of evidence that they had supplied spyware to foreign governments that used it maliciously to target officials, journalists, businesspeople, activists, academics and embassy workers. [22]
In 2024, a group of 23 governments, including the United States, the United Kingdom and several EU member states, explicitly stated in a joint declaration that commercial spyware had been misused in authoritarian regimes and democracies alike. [23]
That admission matters. It means Western governments themselves can no longer sustain the simple division in which political spyware is something done only by “others”.
CIA black sites: when the exception becomes more than surveillance
The darkest case begins where surveillance becomes physical power over a person. After the attacks of 11 September 2001, the CIA established a secret detention and interrogation programme and used secret facilities outside the United States.
The U.S. Senate Select Committee on Intelligence reviewed more than six million pages of material. The final study exceeded 6,700 pages and examined the detention and interrogation of 119 CIA detainees; in December 2014, a 683-page public document was released containing the executive summary, findings and both majority and minority views. [24] [25]
These are not allegations from a foreign government. They are extensive documentation produced by the U.S. Senate.
Europe was not merely an observer
The CIA programme was not only an American story. The European Court of Human Rights found European states responsible for involvement in secret detention.
In Abu Zubaydah v. Lithuania, the Court found several violations of the Convention in 2018, including as a result of Lithuania’s involvement in CIA conduct that led to ill-treatment and arbitrary detention in a secret facility. [26]
In Al Nashiri v. Romania, the Court in the same year examined Romania’s involvement in the CIA secret-detention programme and found multiple violations of the Convention. [27]
And the story did not remain confined to old judgments. On 7 July 2026, in Al Nashiri v. Lithuania, the Court found that al-Nashiri had been held for more than five months in a CIA secret facility in Lithuania in 2005–2006 and that Lithuanian authorities had assisted in his transfer despite a real risk of a flagrant denial of a fair trial and the death penalty in the United States. [28]
This is an especially difficult case of selective universalism. European states were bound by prohibitions on torture, arbitrary detention and secret deprivation of liberty.
Yet when the security exception became politically powerful enough after 11 September, individual Western allies participated in a programme that their own judicial institutions later found to have violated fundamental rights.
The response came only after the secret-detention programme
In January 2009, President Barack Obama issued Executive Order 13491, directing the CIA to close its then-existing detention facilities as soon as practicable and barring it from operating such facilities in the future. For interrogations of persons in U.S. custody or under effective U.S. control in armed conflict, the order required compliance with Common Article 3 and techniques listed in Army Field Manual 2-22.3. [29]
The Senate later released a substantial portion of the findings from its investigation. European courts issued judgments against allied states.
The later disclosures, executive order, Senate investigation and European court judgments show that mechanisms of correction existed.
They do not change the central fact: Western states first created, enabled or tolerated the programme. First came the exception. Only then came the correction.
The strongest Western argument is also the strongest test
Western governments can respond: our intelligence services operate under law; our programmes are reviewed by courts; parliaments can restrict them; journalists can expose them; civil-society organisations can challenge them; and when the system fails, reform is possible. That is a real and important argument. But then it has to apply consistently.
If the measure of democratic legitimacy is independent oversight, necessity, proportionality, time limitation and an effective remedy, those elements cannot be treated as decorative additions around the phrase “national security”. They are the core of legitimacy.
And if we require other states to prove strictly the necessity and proportionality of their surveillance, then the security explanations offered by our own governments cannot automatically be converted into excuses. They should be understood. Then tested by the same standard.
Human rights are not only citizens’ rights
Section 702 raises an even harder question. Under its own constitutional law, a state may provide different procedural rights to its citizens or persons under its jurisdiction than to foreigners abroad.
That is a legal reality. The U.S. Constitution is not a world constitution. But Western political language often speaks not only about American or European civil rights. It speaks about human rights and privacy as a universal value. That creates a consistency test:
would we regard a similar legal and technical logic as acceptable if another major power used it to conduct large-scale interception or targeted collection of our own citizens’ communications?
With Section 702, the comparison has to remain precise: the programme targets foreign persons outside the United States, not the U.S. population as a whole. But the universalist claim remains.
If privacy is a human right, foreign nationality cannot by itself be the end of the moral argument, even if it remains highly relevant to legal jurisdiction.
Democracy is not measured by whether it has emergency powers
Almost every state has emergency powers. The difference lies in what happens to them after the crisis. Do they become ordinary? Do they expand into ever broader categories of threat? Does a temporary measure become infrastructure? Can an oversight body genuinely say no? Will the public even learn that abuse occurred? Will the affected person have a legal remedy? And most importantly:
will we give our own security arguments greater moral weight than comparable arguments made by other states simply because we believe our intentions are better?
Safeguards do not erase the practices they had to correct
Western systems have institutional safeguards. Their existence must not become exculpatory evidence for conduct that those same institutions later had to restrict, invalidate or condemn.
The PCLOB challenged the legal basis and usefulness of the old Section 215 programme. Congress changed it. The NSA deleted improperly acquired data and later abandoned the CDR programme. The FBI sharply restricted Section 702 queries. The FISA Court documented and supervised improving compliance. The ECtHR condemned deficiencies in the old British regime. Britain changed its legislation. The Court of Justice of the EU constrained indiscriminate retention. The European Parliament investigated spyware. The United States restricted companies in the spyware industry. The U.S. Senate documented the CIA programme. The ECtHR found allied European states responsible. That is genuine self-correction. But the list says something else as well.
These institutions were correcting practices that Western democracies had previously authorised, used, concealed or tolerated.
That is not an argument that democracies are equivalent to dictatorships. It is an argument that democracy is not a permanent moral licence.
It is a system for constraining power — and it has to function even when power speaks in the language of security.
How much surveillance, then, can a democracy tolerate?
There is no single number. No single percentage of intercepted communications. No magical category of judicial warrant. But there is a boundary.
A democracy begins to lose its argument when a security exception becomes permanent infrastructure, when intrusion exceeds demonstrable need, when oversight of the overseers is too weak, and when the state asks the public to trust it primarily because the public cannot see what it is doing.
Western systems have an important advantage: mechanisms capable of exposing power, constraining it and sometimes forcing it to retreat. That should be acknowledged. It should not be transformed into proof of automatic innocence.
If the West wants to present privacy, the rule of law and human rights as universal principles, it must accept the consequence of that claim:
“national security” is not a universal exception only when our own government invokes it.
And democracy is not proven by possessing surveillance powers. It is proven by the ability to tell those powers:
enough.
Sources and further reading
- PCLOB — report on the Section 215 telephone records programme Source
- US Department of Justice and ODNI — joint statement on declassification of surveillance documents Source
- NSA — announcement on deletion of call records Source
- US Congress — hearing on the call detail records programme (2019) Source
- ODNI — targeting under FISA Section 702 Source
- PCLOB — Section 702 report (2023) Source
- ODNI — annual statistical transparency report for 2024 Source
- ODNI — annual statistical transparency report for 2025 Source
- ODNI — release on March 2025 Section 702 certifications Source
- PCLOB — Section 702 oversight project Source
- Reuters — explainer on FISA Section 702 (9 June 2026) Source
- Brennan Center — 2026 resource on FISA Section 702 Source
- Brennan Center — analysis of the duration of Section 702 authorities Source
- ECtHR — Big Brother Watch and Others v. the United Kingdom Source
- United Kingdom — Investigatory Powers Act 2016, as enacted Source
- United Kingdom — Investigatory Powers Act 2016, text as at 30 September 2026 Source
- Council of Europe — execution record of the Big Brother Watch judgment Source
- Court of Justice of the EU — press release on data retention judgments (2020) Source
- European Parliament — MEPs’ warning on spyware and democracy Source
- European Parliament — call for investigations and safeguards on spyware Source
- European Parliament — report A9-0189/2023 on spyware Source
- US Department of Commerce — addition of NSO Group and other companies to the Entity List Source
- US State Department — joint statement on countering commercial spyware misuse Source
- US Senate Select Committee on Intelligence — report covering 2013–2015 Source
- US Senate — release of the study of the CIA detention and interrogation programme Source
- ECtHR — Abu Zubaydah v. Lithuania (2018) Source
- ECtHR — Al Nashiri v. Romania (2018) Source
- ECtHR — Al Nashiri v. Lithuania Source
- White House — executive order on ensuring lawful interrogations Source