The Glass Panopticon: The Ethics of Being Watched

Bentham's design was never realized in his lifetime, but the hero image above shows that it was realized eventually, and faithfully. The Presidio Modelo, on an island off the coast of Cuba, was built between 1926 and 1931 directly on Bentham's panopticon plan: four vast circular cell blocks, each six stories of cells ringing a central watchtower, engineered so that a tiny guard staff could observe thousands of prisoners. Its most famous inmates were Fidel Castro and his fellow rebels, held there from 1953 to 1955 after their failed assault on the Moncada Barracks; after the revolution the new government used it in turn to hold its own political prisoners, several thousand at its peak, until it closed as a prison in 1967. Standing empty now, it is the single most literal picture of Bentham's idea in existence: the geometry of total visibility, poured in concrete. But the reason the panopticon matters today is not that a few prisons were built this way. It is that the same geometry, once it no longer needed walls or a tower, spread everywhere.

The tower dissolved into the network. The clearest proof of how far arrived in June 2013, when the contractor Edward Snowden leaked classified documents revealing the scale of US mass surveillance. The first stories showed that the National Security Agency was collecting the telephone metadata of millions of ordinary Americans in bulk, under Section 215 of the USA PATRIOT Act, and accessing user data held by major internet companies through a program called PRISM. The disclosures set off a global debate and led, two years later, to the USA FREEDOM Act of 2015, which curtailed the bulk domestic metadata program. (What Snowden revealed is this article's subject; the far harder question of his own choice to reveal it, and what it cost him, belongs to its own story.) Alongside state surveillance runs the commercial and physical kind: networks of cameras on every corner, and facial-recognition systems that, as the researchers Joy Buolamwini and Timnit Gebru showed in 2018, have been far less accurate on darker-skinned faces than lighter ones, a bias with its own serious consequences that a companion article takes up in full. The point here is structural. Bentham's single watchtower has become a distributed mesh of lenses and servers, and it no longer needs a center to produce the same effect.

Bentham's psychological claim is no longer just a theory; it has been measured. The legal scholar Jon Penney found that in the months after Snowden's June 2013 revelations, views of Wikipedia articles on sensitive topics such as terrorism dropped by roughly 20 to 30 percent, evidence of a 'chilling effect' in which people quietly censor even their curiosity once they know they might be watched, not because they have done anything wrong but because they cannot be sure how a record will one day be read. This is why the common retort, 'if you have nothing to hide, you have nothing to fear,' misunderstands the problem. The legal scholar Daniel Solove has argued that privacy harms are not limited to concealing wrongdoing: the aggregation of innocent data points, the loss of control over how you are seen, and the sheer imbalance of power between watcher and watched are harms in themselves. The philosopher Helen Nissenbaum reframes privacy not as secrecy but as 'contextual integrity,' the expectation that information shared in one context, a doctor's office, a friendship, a search bar, flows only by that context's own norms, so that a violation occurs when data leaks across contexts even if no single fact was ever secret. Europe's General Data Protection Regulation, in force since 2018, tries to give these intuitions legal teeth, with rights to access and erase one's data and fines reaching into the tens of millions of euros or a percentage of global turnover, whichever is greater.

No example is more cited, or more misunderstood, than China's 'social credit system,' and getting it right matters both for fairness and for clear thinking. The popular Western image is of a single, unified number, algorithmically calculated and stamped on every citizen, that rises and falls with their obedience. That picture is substantially wrong. As the corpus documents and China scholars at ChinaFile and the Mercator Institute confirm, the real system, piloted from 2014, is a fragmented patchwork of separate municipal and sectoral programs, many since scaled back, overlapping with a more centrally coordinated set of blacklists aimed mainly at businesses and at people who defy court judgments. Those blacklists are real and serious: by 2019 they had been used to block some 26 million purchases of plane and high-speed-train tickets. But there is no one master score for every person. The accurate account is less cinematic and more useful than the myth, because the myth lets Western readers file the whole problem under 'something authoritarian regimes do,' when the harder truth is that fragmented, purpose-specific scoring by governments and companies alike is spreading, quietly, almost everywhere.

The most ambitious critique of all comes from the scholar Shoshana Zuboff, who named the concept 'surveillance capitalism' in a 2015 essay and developed it at book length in 2019. Her argument is that the largest technology companies have built an economic order on the extraction of human behavioral data as free raw material, to be refined into predictions of what we will do and sold to those who want to influence it, an order she considers a genuinely new and autonomy-eroding form of capitalism. It is a powerful and widely cited thesis, and it is also genuinely contested: whether this is a true rupture in economic logic, or the old business of advertising and marketing operating at unprecedented technical scale, is an open scholarly question this article does not pretend to settle. What can be said is that the raw material of the argument, the pervasive commercial harvesting of behavioral data, is real and documented, whatever one concludes about its ultimate nature.

Three confident claims deserve to be refused, and they cut in more than one direction. The first is the paranoid overclaim: that governments already monitor all digital communication in real time, a single all-seeing eye. They do not; mass-collection programs gather enormous volumes of data but face real limits of storage, processing, and analysis, and genuinely targeted surveillance of specific people remains far more common than any true total-population watch. The second is the specific myth just corrected, that China runs one unified score for every citizen; the reality is a fragmented patchwork, and repeating the myth both misdescribes China and lets everyone else off the hook. But the third overclaim runs the opposite way and is just as unearned: that any and all data collection is inherently tyranny, full stop. That treats a genuine, unresolved tradeoff as already decided. Serious people hold that some surveillance, bounded by oversight, proportionality, and purpose limits, genuinely prevents harm, that privacy norms themselves vary across cultures, and that the real work is regulation rather than either total watching or total blindness. The honest position refuses the security hawk's certainty and the abolitionist's alike. What is not in doubt is Bentham's original insight: that a structure which merely might be watching already changes how free people live, which is why this is a question worth getting right.