Who Actually Governs AI: Part 4 - Surveillance

Clearview AI's website has a page called "Principles." It says the company's facial recognition software exists to protect children from exploitation, help solve violent crime, and promote justice. It reads like a nonprofit's mission statement.

In September 2025, the company's biggest customer for that mission was Immigration and Customs Enforcement. ICE agreed to pay up to $9.2 million for Clearview's search tool, with $3.75 million committed immediately. Weeks later, the Army's special operations command renewed its own multi-year Clearview contract, and Customs and Border Protection signed on for "tactical targeting."

The ethics page and the client list describe two different companies. Only one of them is real, and it is the one writing the invoices.

A Search Engine Built From Everyone's Face

Clearview AI does not photograph anyone. It built a search engine.

The company scraped billions of public images (from news sites, mugshot databases, and social media) and turned them into a matching tool. Upload one photo of a face, and it returns other photos of the same person pulled from across the open internet. The company has said its database now holds more than 70 billion images, and it markets access to more than 3,100 law enforcement agencies across the United States.

Think of it as a reverse image search that only works on human faces, run at a scale no single police department could ever build on its own. That scale is the entire product. It is also the entire controversy.

The Federal Money Kept Coming After the Scandal, Not Before

A reasonable person might assume a company fined tens of millions of dollars for illegal data collection would find government doors closing. The opposite happened.

In September 2025, ICE's Homeland Security Investigations unit finalized its $9.2 million Clearview contract, officially scoped to child sexual exploitation cases and assaults on federal agents. Five months later, Customs and Border Protection signed a separate one-year, $225,000 deal for "tactical targeting" and counter-network analysis. Around the same time, the Department of Justice began onboarding a different facial recognition vendor, ROC, under a contract that could reach $64.3 million over eight years.

Immigration-policy researchers have flagged a real concern here: a tool licensed for narrow investigative purposes sits inside an agency whose broader mission is immigration enforcement, and licensing scope is not the same as usage discipline once the software is installed.

Did You Know?

In March 2026, the Army's 1st Special Forces Command at Fort Bragg renewed its Clearview AI contract through 2030, in a sole-source award that named Clearview the only vendor capable of meeting its needs. The contract requires access to roughly 50 billion images and a 98% accuracy threshold, and the Army's own justification ties the software directly to identifying "high-value targets" early in a military targeting cycle.

New Leadership, Same Product, Different Customer List

Clearview's own leadership changed shape to match where the money was going.

Co-founder Hoan Ton-That stepped down as CEO in December 2024 and left the board entirely by spring 2025. His replacement was Hal Lambert, a Republican donor and investor with existing relationships across the Department of Defense and the Department of Homeland Security. Lambert has said publicly that expanding federal contracts is the company's central growth strategy, with a goal of roughly tripling revenue from an estimated $16 million a year.

That is not a company retreating from controversy. It is a company reorganizing around the part of the market still willing to buy.

A Record-Setting Settlement That Didn't Change the Business Model

In March 2025, a federal judge approved what plaintiffs' attorneys called the largest biometric privacy settlement on record: $51.75 million, structured not as a cash payment but as a 23% equity stake in Clearview AI handed to a nationwide class of people whose faces had been scraped without consent.

The year before, Dutch regulators fined the company 30.5 million euros, calling its face database "illegal." Clearview has said it holds no assets in the Netherlands and does not intend to pay.

Two rulings, two continents, tens of millions of dollars in penalties, and the contract pipeline into federal law enforcement and the military only grew larger afterward. A parking ticket that costs less than the profit from the trip it interrupted does not change anyone's driving.

When the Algorithm Is Wrong, a Person Sits in Jail

Federal testing by the National Institute of Standards and Technology found that some commercial facial recognition algorithms misidentified Asian and Black faces up to 100 times more often than white male faces, depending on the specific system and database used.

That is not an abstract error rate. Randal Reid spent six days in a Georgia jail after being misidentified by facial recognition software he had nothing to do with; the sheriff's office involved later paid $200,000 to settle the resulting lawsuit. In Jacksonville, Jalil Richardson spent more than fifty days in custody after being misidentified as a vehicle thief. Civil liberties groups tracking known wrongful arrest cases tied to facial recognition say seven of the eight people involved were Black.

Clearview AI is not the only company in this industry whose internal ethics structure has been overridden by commercial pressure. In 2022, nine members of Axon's AI ethics advisory board resigned after the Taser maker announced plans for a drone that could fire a Taser at students during a school shooting, a plan the board had already voted 8 to 4 against. Axon paused the project days later, but only after the resignations became public.

The Camera Network Nobody Voted On

Not every gap between principle and practice starts with a press release. In New Orleans, it started with a private camera network.

Since 2023, hundreds of cameras run by a nonprofit called Project NOLA have fed real-time facial recognition alerts directly to New Orleans police officers' phones, flagging matches to a watchlist as people walk down the street. City law allows police to use facial recognition as an investigative tool after the fact. It does not authorize a live, continuous scan of the public.

The ACLU says the program ran for months without City Council approval and, based on the evidence available, appears to still be running. No company's ethics page anticipates a client deciding the rules do not apply to it. That is precisely what makes a written principle different from an enforced one.

Knowlegic Perspective

The pattern across this industry is consistent enough to stop calling it a coincidence. A company writes a mission statement built around protecting the vulnerable. Regulators fine it. A class action settles. Coverage calls it a reckoning. And the government contracts, the actual revenue, keep growing through all of it, often accelerating right after.

That is not hypocrisy in the simple sense. Clearview's leadership never promised to turn down federal money; the "Principles" page was never a legal constraint, just marketing copy sitting next to a sales pipeline. The people who could have made it binding, an ethics board with real veto power, are exactly the people Axon's leadership overrode the moment a board's answer became inconvenient.

What actually governs a surveillance-AI company is not its stated values. It is who is willing to sign the contract, and what happens, or doesn't, when the match is wrong. Facial recognition vendors build their public image around child safety and victim protection. Their fastest-growing customers are the immigration, border, and military agencies that never signed onto those limits.

Read Next: Who Actually Governs AI: Part 5 - Dissent

Sources & References

• ICE Awards Clearview AI $9.2M Facial Recognition Contract, Biometric Update (2025)
• ICE To Pay Up To $10 Million For Clearview Facial Recognition To Investigate Agent Assaults, Forbes (2025)
• US Army Renews Clearview AI Facial Recognition Contract for Special Operations, Biometric Update (2026)
• CBP to Strengthen 'Tactical Targeting,' 'Counter-Network Analysis' With Clearview AI, FedScoop (2026)
• Clearview AI's CEO Resigns As Facial Recognition Company Focuses On Trump 'Opportunities', Forbes (2025)
• Clearview AI, Controversial Facial Recognition Firm, Fined $33 Million For 'Illegal Database', Forbes (2024)
• Wrongful Arrest in US Linked to Facial Recognition Error Leads to $200K Settlement, Biometric Update (2025)