Photo by Lianhao Qu

There is a chapter in Edward Snowden’s memoir Permanent Record that receives less attention than it deserves—not the chapter about PRISM or XKeyscore or the Hong Kong hotel room, but the chapter describing what his colleagues actually did with all that power when nobody was watching, which, in a surveillance state, nobody ever is.

The program called XKeyscore is best understood as a search engine for the entirety of a person’s digital life. Type in a phone number, an email address, or an IP address, and you can read the emails, listen to the calls, and watch the browsing history unspool in real time—search queries appearing letter by letter as they were typed, what Snowden calls, with bleak precision, “a human complete.” The system was so comprehensive that Snowden realized he could have searched the communications of a Supreme Court justice or a sitting member of Congress with negligible risk of detection.

What his colleagues actually used this power for—what they found the system most reliably useful for in their day-to-day practice—was watching people they wanted to sleep with.

The practice had a name. The intelligence community, with its compulsive bureaucratic instinct to acronymize everything it touches, called it LOVEINT—a deliberate rhyme on HUMINT and SIGINT and, as Snowden notes, a gross joke on both. Analysts used XKeyscore, phone intercepts, camera feeds, everything, to monitor current lovers, ex-lovers, and objects of more casual affection. They read emails. They listened to calls. They looked at photographs. The NSA acknowledged the practice when pressed and characterized it as individual misconduct. It was nothing of the sort. It was endemic enough to have a name, widespread enough that the analysts Snowden worked alongside discussed it without apparent embarrassment, and protected by the same structural logic that protected every other abuse of the system: to prosecute an analyst for LOVEINT would require acknowledging in open court that the system being abused existed. The government would never do that. The analysts knew the government would never do that. The knowledge was shared, and the knowledge was itself a form of institutional permission.

“Intercepted nudes were a kind of informal office currency,” Snowden writes. The unspoken rule was that if you found a naked photograph or video of an attractive target, you showed the rest of the team—at least when there were no women present, a caveat that speaks volumes about how rare women were in these rooms. “That was how you knew you could trust each other: you had shared in one another’s crimes.”

The atmosphere Snowden describes in vault V22 at NSA headquarters — decorated, let the detail stand, with a seven-foot poster of Chewbacca — is not the atmosphere of men who believe they are doing something wrong. It is the atmosphere of men who believe they are entitled. The technology did not corrupt them. It handed them a key.

This is the surveillance state at its most human, which is to say, its most revealing. The apparatus built to protect a democracy from its enemies was being used by young men to look at photographs of women who had no idea they were being watched. The technology did not create the entitlement. It removed the friction that previously constrained it.

Snowden describes the intelligence community of that era as crowded with men—technically credentialed, often very young, and operating within a culture that had developed a specific ideology to justify itself: the conviction that the analysts knew better than the citizens they were surveilling, that democracy’s ordinary protections were obstacles for the naive, and that the contractor badge conferred a kind of moral authority. XKeyscore was also a game—gamified with competitive logic that rewarded the most aggressive use. The citizen whose communications were being parsed was not a rights-bearing individual under the Fourth Amendment. They were content. Data. A score.

But the logic that made it possible — the conviction that the inner life of others is legitimately available to those with the technology to access it — does not stay in the vault. It travels. It goes private. It goes global. It gets productized, exported, and sold to governments that ask no questions and face no consequences. And it goes looking for deeper frontiers.

The NSO Group was founded by veterans of Israeli military intelligence Unit 8200. Pegasus, their flagship product, is a zero-click exploit: no link is clicked, no attachment is opened—a message arrives and the device is compromised, everything on it available to the operator. NSO marketed it as a counterterrorism tool. The gap between that description and its documented uses tells the story of the apparatus in miniature.

Saud al-Qahtani, senior adviser to Mohammed bin Salman, ran a troll army of Twitter bots and operators dubbed “the flies” by Saudi dissidents—their purpose to swarm critics, amplify regime propaganda, and manufacture the appearance of popular support for a government with no legitimate means of generating it. Before tuning in via Skype to oversee the murder and dismemberment of journalist Jamal Khashoggi, al-Qahtani’s portfolio included running social media operations for the royal court. Khashoggi and Montreal-based dissident Omar Abdulaziz built a counter-network they called the Bees, designed to amplify independent voices, the flies were working to drown out. By 2018, the Bees were gaining traction. In 2018, Khashoggi was asked to visit the Saudi consulate in Istanbul for routine paperwork for his upcoming marriage. He entered on October 2nd. He did not come out. Whole.

After his death, investigators discovered that Pegasus had compromised Abdulaziz’s phone before the murder. The communications between them — their planning for the Bees, their discussions of the consulate visit, their assessments of MBS’s likely responses — had been available to Saudi intelligence in real time. The flies had been watching the Bees all along. Monitoring, swarming, and killing were not separate departments. They were consecutive stages of the same process.

A friend of Khashoggi’s described what those mornings were like: waking up to “the equivalent of sustained gunfire online.” Friends made a standing habit of calling to check on his mental state. This was not unpleasant press coverage. This was a coordinated state operation running through a platform built in California.

The case of Jeff Bezos adds a dimension the Khashoggi murder might obscure: the apparatus is indifferent to power and wealth. In May 2018, Bezos received a WhatsApp video file from an account belonging to MBS. The two men had exchanged numbers at a Hollywood dinner. UN Special Rapporteurs Agnes Callamard and David Kaye assessed with medium-to-high confidence that the video delivered malware, with massive data extraction following within hours. Nine months later the National Enquirer — whose parent company has documented ties to the Saudi government — published intimate details of Bezos’s private life. Bezos publicly identified this as extortion linked to Saudi Arabia’s desire to silence the Washington Post‘s coverage of Khashoggi’s murder. A head of state surveilled the richest person on earth using commercially available spyware and then blackmailed him with the proceeds. The apparatus does not respect the hierarchy it surveils. It uses the hierarchy against itself.

The Khashoggi name connects these threads across generations. Adnan Khashoggi — Jamal’s first cousin, one generation older — brokered billions in arms deals between American defense contractors and the Saudi military, serving as middleman in both the BAE/Al-Yamamah deal and Iran-Contra. The UK Serious Fraud Office investigation into Al-Yamamah was shut down in 2006 by Prime Minister Tony Blair on national security grounds — a British court found the decision unlawful. The SFO subsequently lost 32,000 pages of investigation documents. Nobody was ever prosecuted anywhere. The structure of impunity that protected Adnan is the same structure in which MBS murdered Jamal and faced no meaningful consequences. The technology updated. The impunity did not.

Twitter’s role in this apparatus was not passive. In 2019 the Department of Justice charged two former Twitter employees with using administrative access to expose Saudi dissidents to the government in exchange for payments exceeding $300,000. One was convicted in 2022. The platform had been actively penetrated. The flies were not just using Twitter. They were, in a meaningful operational sense, inside it. Twitter — rebranded X — is now owned by Elon Musk, who also co-founded Neuralink. The man who owns the platform the Saudis infiltrated to track dissidents is building a device designed to read the thoughts of its users. The through-line is not coincidence. It is infrastructure.

The Israeli government used Pegasus as diplomatic currency. Nearly every signatory of the 2020 Abraham Accords subsequently received access to the tool — the UAE had been a Pegasus client since 2013, seven years before formal normalization, in what activists have named Pegasus diplomacy. The Abu Dhabi cybersecurity firm DarkMatter recruited Unit 8200 veterans and former CIA and NSA officials with million-dollar packages, later exposed as the developer of ToTok—a messaging app that functioned as a covert mass surveillance tool, harvesting communications data from users who believed they were using an ordinary messaging service. The mad scientist of popular imagination is a distraction from this actual structure. The scientists are not insane. They are recruited, well paid, and working in purpose-built facilities in cities designed to attract international talent in jurisdictions where what they build can never be investigated or challenged. The accountability vacuum was engineered.

At the outer edge of this logic sits Havana Syndrome—the cluster of neurological symptoms experienced by hundreds of American diplomatic and intelligence personnel at embassies worldwide since 2016, beginning in Cuba and spreading to Vienna, Beijing, and beyond: sudden-onset auditory phenomena, vertigo, cognitive deficits, and in some cases lasting brain injury. The National Intelligence Council’s December 2024 assessment holds it “very unlikely” a foreign actor deployed a novel weapon, though two of seven agencies dissented. By pointing consistently away from Russia and China — the named adversaries — the assessment structurally protects a third category: Gulf state allies whose intelligence services have demonstrated both the will and the means to deploy sophisticated technological weapons against individuals. The structural parallel to Blair’s termination of the BAE/Al-Yamamah investigation is precise: in both cases, the official conclusion protects the alliance. James Giordano, a neuroweapons expert directly involved in the Havana Syndrome investigation, confirmed publicly that the United States possesses weapons capable of producing the reported symptoms. The question of who deployed them against American personnel remains officially open. The accountability vacuum has teeth.

The FBI purchased Pegasus in 2019. NSO created a US-specific version called Phantom capable of targeting American phone numbers — standard Pegasus could not, per Israeli export license conditions. The bureau spent two years deliberating whether to deploy it domestically. In November 2021, the Biden administration publicly blacklisted NSO Group, citing evidence the tools had been used against “government officials, journalists, businesspeople, activists, academics, and embassy workers.” In the same month, a secret contract was signed to purchase Pegasus again, routed through a shell company called Cleopatra Holdings—actually a New Jersey contractor named Riva Networks, used for the same purpose two years earlier. When the contract became known, the White House directed the FBI to investigate. The FBI investigated and found the client was itself. The bureau stated it had been misled and used the tool “unwittingly.”

Public legitimacy gets maintained through condemnation; operational capability gets maintained through procurement; a shell company named after an Egyptian queen absorbs the accountability gap. The scandal is that there is no scandal at all.

Edward Snowden understood the system from the inside. “I had hoped to serve my country,” he writes in the chapter he calls “Homo Contractus.” “But instead I went to work for it. This is not a trivial distinction.” He wore a Dell badge to work at Langley. He was working for the CIA. Government capability, private employment, public money, no public scrutiny — the contractor badge is the solution to the problem of democratic accountability, worn by the people who designed the problem.

In 2018, The Intercept revealed Google’s Project Dragonfly: a censored search engine developed secretly for China, designed to link search queries to phone numbers and block results relating to democracy, human rights, peaceful protest, and religious belief. Google had no intention of disclosing its existence. More than 200 employees signed an open letter calling it a betrayal, linking it explicitly to Project Maven—Google’s Pentagon contract for drone targeting AI—as twin expressions of the same product logic: the market for surveillance tools does not distinguish between authoritarian and democratic clients. It distinguishes between paying and non-paying ones.

The same logic that built Dragonfly for Beijing built Rekognition for ICE and built the data pipelines that Meta’s AI chatbot now runs on. The engineers who built Dragonfly did not unlearn what they learned when the project was officially terminated. The techniques developed for the Chinese censorship market—linking behavioral data to individual identities at scale, suppressing specific content categories on government instruction — are not architecture that disappears when a project is cancelled. They are knowledge, and knowledge does not have a national address. Three accountability failures converge in the Dragonfly episode: a private company conducting what amounted to a foreign intelligence partnership without democratic mandate, a destination jurisdiction where the product could never be challenged by those it would target, and a company operating with sovereign-adjacent reach and private-sector immunity. The East India Company extracted resources from territories it did not govern under a royal charter that gave it sovereignty-adjacent power without sovereignty-adjacent accountability. We are living in its software sequel. In December 2025, Meta began using chatbot interactions to personalize advertising. A coalition of thirty-six privacy organizations described this to the FTC as “a deliberate strategy to normalize a fundamental expansion of surveillance-driven and behavior-changing marketing.” Linguist Emily Bender put the structural logic directly: “Before, Meta’s systems watched who you connected to and what your communities were doing. Now it’s direct: What are you saying to the company?”

The experience of typing into a chatbot feels like a private conversation—intimate and, in some cases, therapeutic. People share things with chatbots that they would not post publicly and would not enter into a search bar, because the conversational register signals safety. That signal is false. One study found Meta’s chatbot collects more user data than any other major AI platform—including DeepSeek, developed under CCP influence. Meta out-surveils the CCP. Chew on that.

The engineers who built the behavioral modification systems of social media sat before the cameras in the 2020 Netflix documentary The Social Dilemma and tried to describe what they had made. They reached for metaphors—marionette strings, Pavlovian bells, laboratory rats pressing levers—and each metaphor failed to contain what they were describing. The discomfort in their faces was the discomfort of people who had built something they could not name. What they were struggling to articulate is the oldest logic in Western history, now turned inward. The Portuguese mapped coastlines. The Spanish followed. The English consolidated. Each successive wave pushed further and went deeper, and the interior—the highlands, the river systems, the human interior—was always the final destination because that is where the most valuable and least defended resources turned out to be.

The final frontier is not outer space. It is the space between your ears.

In November 2018 — the same month the world was processing Khashoggi’s murder—Chinese scientist He Jiankui announced the birth of the world’s first gene-edited babies, whose germline DNA had been altered using CRISPR-Cas9, ostensibly to confer HIV resistance. He had studied at Rice University and Stanford. He had told American geneticists there his intentions before proceeding. They expressed disapproval. Nobody stopped him. He was convicted and sentenced to three years in prison by a Chinese court. But what condemnation could not undo was the fact. The germline had been edited in a living human who would later have children. The threshold had been crossed. The genie, to use Snowden’s own metaphor for the surveillance apparatus he exposed, was out of the bottle.

What matters for this argument is not what He Jiankui claimed but what his experiment proved possible. The gap between “HIV immunity” and “enhanced cognition” is not a technical gap. It is a question of which genes you target and who funds the research. He crossed the threshold in a jurisdiction chosen partly for its regulatory permissiveness, with the prior knowledge of researchers in jurisdictions with stronger oversight who chose not to intervene. The capability, once demonstrated, cannot be uninvented. The question of what it will next be used for, by whom, and in which jurisdiction chosen for its distance from oversight—nobody in a position to answer it has any incentive to ask it.

DARPA’s brain-computer interface programs have followed the same trajectory: framed as therapeutic for injured soldiers, with enhancement and external control ambitions legible in the funding history. The technology migrates with its assumptions. The internet carried DARPA’s assumptions about distributed communication into civilian infrastructure. GPS carried military tracking assumptions into the pocket of every smartphone user on earth. The BCI carries assumptions about cognitive access and control into the neural interface market, where they are currently being commercialized under therapeutic framing, structurally identical to every other humanitarian alibi in this story. Cure the paralyzed. Restore the injured. Then see what else the device can do.

Elon Musk has described Neuralink’s goal as “digital telepathy”—the direct merging of human cognition with AI systems. He owns Starlink (communications infrastructure across roughly a hundred countries); X/Twitter (the platform Saudi intelligence infiltrated); xAI/Grok (trained on X’s data—the accumulated expression of billions of people’s opinions, emotions, and private disclosures); and Neuralink, which aspires to read the neural signals that precede that expression. The convergence of communications infrastructure, information platform, AI development, and neural interface in a single unelected private actor may or may not reflect malicious intent. Democratic governance has not caught up either way.

Once surveillance capitalism becomes the dominant business model, competition does not restrain the apparatus. It accelerates it. Each actor must out-surveil its competitors to capture behavioral data more intimately and continuously, enabling the sale of more precisely targeted behavioral modification to advertisers and states. The ratchet turns one way. What was considered outrageously intrusive in 2013—XKeyscore, mass metadata collection at population scale—is now the assumed baseline of every free application on every networked device. The floor has dropped so far that users have stopped noticing it move.

This explains the central puzzle: why does the apparatus keep growing despite continuous public exposure? Snowden revealed everything in 2013. The apparatus grew. The Pegasus Project exposed systematic spyware abuse in 2021. The client list expanded. Dragonfly was killed by an employee revolt. The same logic that built Rekognition for ICE, built the Meta chatbot pipeline, and is building Neuralink. Public exposure does not stop the apparatus because the incentive structure outlasts the outrage. Shame gets priced in, absorbed, and the expansion continues.

The terminus of this dynamic is what might be called nano-totalitarianism: control distributed so finely through every interaction, every interface, every intimate chatbot disclosure, that there is no single point of resistance because there is no single point of control. Not Orwell’s boot stamping on a face—too crude, too visible, too resistible. The panopticon is internalized until the prisoner surveils themselves. Total without being obvious. And if the neural interface arrives at scale — if the chatbot that harvests what you say gives way to the device that harvests what you think before you decide to say it — the permanent record becomes genuinely complete. The ledger moves from what you did to what you said to what you thought before you said it.

The conquistadors did not announce conquest. They arrived with mirrors and beads. The search engine that organizes the world’s information. The social network that connects you to everyone you have ever known. The intimate chatbot that listens without judgment. The neural interface that restores movement to a paralyzed hand. Each one a mirror—reflecting back a version of yourself that felt like recognition. Each one is a bead—something shiny exchanged for something of incalculably greater value. The exchange seemed weightless: a little data for a lot of convenience. By the time the terms of exchange become visible, the ledger is already written in someone else’s favor.

The permanent record — Snowden’s phrase and his title, chosen with precision — is the colonial ledger of the digital frontier. XKeyscore metadata. Pegasus-extracted communications. Behavioral profiles assembled from a decade of social media. Chatbot confessions volunteered in the belief that no one was listening. Neural signals, in the near future, will be recorded before they become words. Cumulative, cross-referenceable, permanent. The subject has no access to it. The subject cannot correct it.

Democratic accountability is the only structural response. Not individual privacy hygiene, though it matters at the margin. Not encryption alone, though it slows the extraction. The political question—who owns the ledger, who can access it, under what legal framework, and who is accountable for its use—is the question the conquistadors never answered because the conquered had no standing to ask it. The British parliament never investigated what the East India Company was doing in Bengal. The Spanish crown never meaningfully constrained the encomienda. Official inquiries were suppressed, documents were lost, and accountability was deferred indefinitely in the name of strategic necessity and alliance politics. Blair killed the BAE investigation. The FBI investigated itself and found itself innocent. The IC assessment exonerated the Gulf allies. The pattern has a name: impunity managed as policy.

In the digital frontier, the conquered still have standing. They have not yet asked the question loudly enough to compel an answer. But the window for asking it closes a little each year—with each new product release, each new data pipeline opened, and each new threshold crossed in a jurisdiction chosen for its distance from democratic accountability.

Snowden ends Permanent Record with a question about what kind of world we want to leave. This piece ends with a simpler one.

Whose record is it?

John Kendall Hawkins is an American ex-pat freelancer based in Perth, Australia. He is a former reporter for the New Bedford Standard-Times.