SeriesHidden Power: Dynasties, Secret Societies and the Invisible Mechanisms of Power Part 49 / 55

The Digital Panopticon

Hidden Power — VII. The Psychology of Power and the Control of Human Beings — SM-49

Editorial illustration of everyday digital traces; it does not depict a specific surveillance system.

Hidden Power — VII. The Psychology of Power and the Control of Human Beings — SM-49

Our everyday digital footprint is made up of small events: looking up directions, watching a video, using a loyalty card, clicking an advertisement, or talking to an acquaintance. Each event may seem insignificant. But when data are linked together, they can produce a portrait of a person that the person never consciously assembled. Who can see that portrait? Who sells it, uses it to make decisions, or uses it to steer our attention? And do we already behave differently simply because we know we might be watched?

The term digital panopticon is appealing because it brings all these concerns together in a single image. Yet it can also mislead: online platforms, data brokers, advertisers, government agencies, and operators of spyware are not one institution. Their capabilities differ, as do their legal powers, motives, and demonstrable consequences. That is precisely why we must investigate the links between them rather than assume those links exist.

The Tower Whose Observer Need Not Be Seen

The idea of the panopticon predates the computer by a long way. As the Bentham Project at University College London explains, Samuel Bentham devised a principle of centralized supervision of work while in Russia. His brother Jeremy adapted it into a proposal for a prison with cells arranged around a central observation tower. The original letters were written in 1786, with later additions in the early 1790s. The central idea was not that a governor would actually watch everyone at every moment, but that those being watched could never know for certain when they were under observation. [1]

Historical black-and-white architectural drawing of the panopticon, showing a multi-storey section above a circular floor plan with cells around a central inspection area.
A panopticon plan published in Bentham’s collected works (1843; original design 1791). The central inspection principle underlies the metaphor; it does not document a unified modern surveillance system. Image: Jeremy Bentham / The Works of Jeremy Bentham, vol. IV (1843) / Wikimedia Commons Public Domain (Wikimedia Commons record; Public Domain Mark, not CC license)

The philosopher Michel Foucault later saw in the panopticon a symbol of disciplinary power: external observation can turn into internal self-surveillance. Yet neither the historical building nor the philosophical interpretation proves that contemporary society operates as a single prison. The metaphor is useful when it prompts us to ask who can observe, how that ability is distributed, and whether those being watched can challenge it. It is not, however, an accurate technical map of the internet. [2]

Modern systems have another important distinction: they do not merely observe behavior; they also use data to calculate probabilistic inferences. The object of surveillance can become a profile predicting a purchase, an interest, a financial risk, or a likely response. Such a profile may be useful, mistaken, or unfair; it is not thereby a true definition of the person.

A Market in Fragments That Form a Profile

In 2014, the US Federal Trade Commission (FTC) examined nine data brokers. It found that they obtained information from commercial and other records, online and offline sources, and other brokers. Seven of the nine exchanged data with at least one other broker in the group studied. One broker examined held information on more than 1.4 billion consumer transactions and 700 billion data elements. These are not figures for 1.4 billion different people; they concern transactions and data elements, and it would be wrong to turn them into a count of individuals being monitored. [3]

What matters here is the circulation of information. A customer may give a piece of information directly to one organization, only for it later to be combined with other databases the customer never considered. The FTC also described inferences about age, income, preferences, and sensitive circumstances. The underlying data often do not establish whether an inference is correct. Someone who buys a book about an illness is not necessarily ill; someone who visits a political website does not necessarily support its views. Yet a mistaken inference may still affect an offer, a ranking, or a risk assessment.

This market also has legitimate purposes, which nevertheless need to be assessed separately from specific collection practices and subsequent uses. Data can help detect fraud, make services easier to use, and provide offers more relevant to users. But those benefits do not remove the questions of whether people know how their data are used, whether they can correct errors, and how long the information is retained. The FTC specifically warned about the industry's poor transparency and individuals' limited ability to exercise control. Its report describes a sample of nine companies at a particular time, not every data broker in the world. [3]

The Platform That Knows What We Never Told It

A separate FTC study, published in September 2024, examined the practices of nine major social-media and video-streaming services. It was based on special information requests issued by the regulator; its findings do not show that every problem occurred at every company to the same extent. The report describes extensive data collection, the possibility of lengthy retention of information about both users and non-users, tracking beyond the service itself, and the use of data for advertising, recommendations, and automated analytics. [4]

A typical example is tracking code on a third-party website. When someone visits that website, certain technological systems can send information to a platform even though the person has not opened the platform's own app at that moment. This does not mean a platform can see every aspect of someone's life or all their conversations without limit. It does mean, however, that a person's data footprint cannot be understood solely from what they deliberately post on their profile.

An important form of influence is not just advertising but the selection of what becomes visible. Recommender systems can rank content using different signals: past interests, relationships, language, response times, or the service's goals. It does not follow that every algorithmic selection is political manipulation. Ranking is also necessary to make large quantities of content usable. The crucial questions are who sets the measure of success and how far users understand and control the effects of that measure.

When a Profile Becomes a Political Weapon

The UK's Information Commissioner's Office (ICO), in its investigation associated with Cambridge Analytica, documented the collection of information about millions of Facebook users through the This Is Your Digital Life app and the subsequent use of data in a political setting. The ICO describes how political actors attempted to use digital data to reach small groups of voters with tailored advertisements. This is evidence of data collection, transfer, and targeting practices; it is not direct proof that Cambridge Analytica determined the outcome of an election or referendum. [5]

A 2023 study published in the Proceedings of the National Academy of Sciences (PNAS) used randomized experiments to examine how much microtargeting might increase persuasive effects. In one of the settings studied, the authors estimated a substantial advantage over less carefully chosen ways of selecting messages, but that advantage depended on the type of messages, the comparison strategy, and the data available. The research result is therefore not proof of a universal machine for shaping political will. [6]

More recent experiments are not unequivocal either. A 2024 study of political persuasion using GPT-4 found no statistically significant overall advantage for individualized microtargeting over comparable non-personalized messages. That does not mean targeted advertising never works; it means effectiveness must be measured under specific conditions rather than inferred merely from a technology's capabilities. [7]

This is the difference between being able to reach a particular audience, being able to tailor a message, and demonstrating that the message actually changed beliefs or behavior. The first two are technical and organizational capabilities. The third is an empirical question. Another open question is what happens when many messages, advertising auctions, social pressures, and recommendations interact outside a controlled experiment.

State Surveillance Is Not the Same as Advertising-Based Profiling

Commercial tracking and state interception can intersect through the same data infrastructure, but they operate under different legal regimes. Section 702 of the US Foreign Intelligence Surveillance Act (FISA) authorized the targeted collection of foreign-intelligence information concerning non-US persons reasonably believed to be outside the United States. Other people's communications could be collected incidentally. The intelligence community's official explanation emphasizes targeting and protective procedures; that account must be read alongside, not in place of, the findings of independent oversight bodies. [8]

A staff report of the US Privacy and Civil Liberties Oversight Board (PCLOB), published in April 2026, describes changes following the 2024 reforms, procedural improvements, and unresolved civil-liberties concerns. Positive assessments in that report do not mean the reforms' long-term impact has been fully measured. An important qualification is that this is a staff report, not a document approved by a quorate board. It also identifies the intelligence-gathering benefits of the program. Research integrity requires acknowledging both: the documented risks to rights and the specific security purposes at issue. [9]

A further legal complication arose in 2026. Following temporary extensions, the statutory authority under Section 702 expired on June 12, but the Congressional Research Service explains that existing judicial authorizations could remain valid until their own expiration. This is neither an indefinite extension of the law nor an immediate halt to all collection. Any subsequent reauthorization would require a separate, up-to-date check before the article is published. [10]

This is precisely where the phrase “all-seeing surveillance” is too imprecise. Investigators need to ask which institution obtained what information, under which authority, who queried the database, and how the legality of that action could be challenged. Without these distinctions, we risk both underestimating real abuses and exaggerating assumptions of omnipresence.

A Targeted Infection Does Not Mean Everyone Is Infected

Sometimes surveillance does not involve gradually assembling profiles but instead directly compromising an individual's device. This is a different technological and evidentiary mechanism from collecting data for advertising. In February 2024, Citizen Lab and its partners reported on at least 35 members of Jordanian civil society who had been targeted by or infected with Pegasus spyware between 2019 and September 2023. Citizen Lab directly verified forensic evidence in 30 cases; evidence concerning five additional cases was analyzed by other partners. The researchers explicitly distinguish confirmed infection from traces of attempted targeting. [11]

That distinction matters. Evidence of an attempted attack does not guarantee that the spyware successfully took over a phone. Conversely, the absence of every digital trace does not necessarily mean an infection never occurred. The strength of such an investigation lies in clearly stating which findings the evidence supports and which it does not.

Cases involving the surveillance of activists, journalists, and lawyers show the possible consequences of access to sensitive messages, contacts, and private circumstances. But documented targeted intrusions do not justify a leap to the claim that a single actor continuously monitors every phone. Nor does a spyware producer's nationality, by itself, establish who selected a target or operated the software in a specific attack.

Rights Exist — But People Need to Be Able to Exercise Them

The European regulatory framework raises another kind of question: not only what is technically possible, but what a data controller is legally permitted to do. The General Data Protection Regulation (GDPR) provides a right of access to personal data in Article 15 and a right to object on certain grounds of processing in Article 21. Article 22 establishes conditions and safeguards concerning decisions based solely on automated processing that produce legal or similarly significant effects. None of these provisions means that all profiling is entirely prohibited or that an individual's request must always be granted without exceptions. [12]

The Digital Services Act (DSA) requires, among other things, transparency about online advertising; it restricts the presentation of ads based on profiling that uses special categories of personal data, and it restricts profiling-based advertising to minors when a provider knows with reasonable certainty that the user is a minor. It also requires very large platforms to offer at least one content-recommendation option that is not based on profiling. These obligations are not a general ban on recommendation algorithms or all online advertising. [13]

In 2021, the European Data Protection Board published guidelines on targeting social-media users. Legal rules therefore also address the responsibilities of platforms and advertisers. Yet the gap between a written rule, its actual enforcement, and a meaningful choice that users can understand still calls for scrutiny. [14]

Individuals can therefore take concrete, if incomplete, steps: review advertising and recommendation settings, limit unnecessary app permissions, keep devices updated, and, where appropriate, exercise rights of access, objection, or rectification. But privacy must not become solely the burden of a person forced to guess where their data have gone. Transparent oversight, effective sanctions, independent journalism, accountable researchers, and the ability to scrutinize institutions publicly are also essential.

Who Is Allowed to Watch Whom?

The question of whether a digital panopticon exists cannot be answered fairly either with casual dismissal or with an image of one hidden manager of the world. There are documented commercial networks for collecting and trading data, algorithmic ranking of information, political targeting, lawful and contested forms of state interception, and forensically verified cases of targeted intrusion. But the evidence does not warrant the conclusion that all these systems share a single command structure, that every device is continuously monitored, or that they can determine the human will without limits.

Perhaps the most important consequence of digital surveillance is less dramatic but more enduring: an asymmetry of knowledge. An institution can assemble profiles of a person that the person cannot see, while that person may struggle to find out why they were categorized, approached, or excluded. Against this stands a principle: power that demands transparency from people must itself be open to scrutiny.

Freedom does not mean that nobody can ever see us. It means observation must not be arbitrary; it must have an explainable purpose and a legal boundary; decisions must be open to challenge; and a human being must remain more than a collection of data-driven predictions. A single privacy setting cannot meet these demands. They require institutions that can be held to account and people who do not surrender their own judgment.

Sources and further reading

  1. University College London, Bentham Project. The Panopticon. source Source
  2. Stanford Encyclopedia of Philosophy. Jeremy Bentham (overview of the historical background and philosophical interpretations). source Source
  3. Federal Trade Commission (May 27, 2014). FTC Recommends Congress Require the Data Broker Industry to be More Transparent and Give Consumers Greater Control Over Their Personal Information; summary of Data Brokers: A Call for Transparency and Accountability. source Source
  4. Federal Trade Commission (2024). A Look Behind the Screens: Examining the Data Practices of Social Media and Video Streaming Services (staff report, September 2024). source Source
  5. Information Commissioner's Office. Cambridge Analytica raids (official retrospective account of the investigation begun in 2017). source Source
  6. Research article (2023), Quantifying the potential persuasive returns to political microtargeting, Proceedings of the National Academy of Sciences 120(25), e2216261120. source Source
  7. Research article (2024), Evaluating the persuasive influence of political microtargeting with large language models, Proceedings of the National Academy of Sciences, DOI 10.1073/pnas.2403116121. source Source
  8. Office of the Director of National Intelligence / Intelligence.gov. Incidental Collection in a Targeted Intelligence Program (the executive branch's official explanation). source Source
  9. Privacy and Civil Liberties Oversight Board (April 2, 2026). Report on the Surveillance Program Operated Pursuant to Section 702 of FISA, staff report (not approved by a quorate board). source Source
  10. Congressional Research Service (June 22, 2026). The Impact of FISA Section 702's Repeal, Legal Sidebar LSB11444 (publicly accessible mirror of the report). source Source
  11. The Citizen Lab (February 1, 2024). Confirming Large-Scale Pegasus Surveillance of Jordan-based Civil Society (with Access Now and partners). source Source
  12. European Parliament and Council of the European Union (2016). Regulation (EU) 2016/679, General Data Protection Regulation (GDPR), particularly Articles 15, 21, and 22. source Source
  13. European Parliament and Council of the European Union (2022). Regulation (EU) 2022/2065, Digital Services Act (DSA), particularly Articles 26 and 38. source Source
  14. European Data Protection Board (April 13, 2021). Guidelines 8/2020 on the targeting of social media users — Final version. source Source