AI surveillance uses AI to watch many people by sorting data about them. It now costs so little that company contract rules are the main limit. This year the US government blacklisted Anthropic for refusing to drop those rules.
Anthropic refused the Pentagon’s terms and was blacklisted
In February, the Pentagon asked to use Anthropic’s AI, Claude, for “all lawful purposes.” Anthropic wanted two exceptions in the contract: no mass surveillance of Americans, and no weapons that fire without a human decision. The Pentagon refused limits on individual uses. Anthropic’s CEO said the company “cannot in good conscience accede.”
The government then ordered federal agencies to stop using Anthropic. It labeled the company a “supply chain risk,” a label usually kept for companies tied to foreign adversaries. A federal appeals court let the label stand in September.
OpenAI signed its own Pentagon deal within days. After public backlash, it banned intentionally using its AI to watch Americans, including through data the government buys. Critics noted that the public would likely learn of a broken ban only through a whistleblower.
AI removes the cost that once limited surveillance
Surveillance once needed people. Watching someone took agents, time, and money, so governments watched only a few people closely. AI reads, sorts, and summarizes data about millions of people at once.
The data already exists. A Director of National Intelligence report, made public in 2023, confirmed that US intelligence agencies buy personal data from brokers. Apps sell the same kind of data to advertisers. The report warned that in the wrong hands this data could enable blackmail, stalking, and harassment. AI turns that pile of data into answers about specific people.
Company policies are the main wall, and policies change
A handful of company policies now stand between that data and mass AI surveillance. The law moves slowly. Congress fought for months this year over renewing Section 702, the law that lets agencies collect communications. It passed short extensions while it argued.
A new contract, a new CEO, or government pressure can change a company policy fast. This year one company held its line and was punished. Another rewrote its deal within days. Whatever your politics, that wall is thin.
Your AI data sits on someone else’s computers
Anything you put into an AI service lives on someone else’s computers, under someone else’s rules. Consumer AI apps keep your chats, and some train on them by default. The government can request stored data, a lawsuit can demand it, and the company can sell or leak it.
Nobody needs to watch you today. The data only needs to exist and the rules around it to change.
Keep sensitive work in AI you control
- Share less with consumer AI apps. Leave out names, client details, and anything you wouldn’t want read back to you.
- Turn off chat history and training where you can. Most AI apps keep both settings under privacy.
- Use AI you control for sensitive work. An open model on a computer you control keeps your conversations away from AI companies.
- Keep your most sensitive conversations with people. A trusted colleague keeps no log on a server.
Bottom line: AI makes mass surveillance cheap, the data to feed it is already for sale, and company rules that bend under pressure are the main limit. Keep sensitive work in AI you control. PageLines runs on open models, and each member gets an AI that works only for them. See how it works.
Common questions
What is AI surveillance?
AI surveillance uses AI to watch many people by sorting data about them, such as location, purchases, and messages. AI makes it cheap because it analyzes data about millions of people at once.
Why did the Pentagon blacklist Anthropic?
Anthropic refused to drop two limits on its AI: no mass surveillance of Americans, and no weapons that fire without a human decision. The government labeled it a supply chain risk. An appeals court let the label stand.
How can I protect my data from AI surveillance?
Share less with consumer AI apps, turn off chat history and training, use AI you control for sensitive work, and keep your most sensitive conversations with people.

