On eve of IPO, Anthropic Founders Warn AI Could Lock In Authoritarian Surveillance States
The company selling tools to governments admits the danger of high-tech repression while heading toward a massive IPO

In the letter that opens Anthropic’s IPO prospectus, the founders warn that authoritarian governments may use AI to create high-tech surveillance states, deepen repression, and make themselves much more difficult to dislodge. All seven co-founders signed the letter, including CEO Dario Amodei and President Daniela Amodei. Authoritarian abuse is listed as one of four broad risk categories, alongside misuse, loss of control, and economic disruption. The company’s stated response is a commitment to work with free societies as a counterweight and to promote freedom and democracy.
The same filing notes that Anthropic already sells AI to governments to build technological state capacity. It works with the U.S. federal government, many state governments, and American allies in areas that include child services, Social Security administration, cybersecurity, and intelligence collection.
Anthropic’s usage policy bars its models from fully autonomous weapons or mass surveillance, including mass domestic surveillance. Holding that line has cost the company revenue. After the position became public, the Trump administration directed federal agencies to stop using Anthropic’s technology. The Pentagon designated the firm a supply-chain risk to national security and raised the possibility of invoking the Defense Production Act to force the company to drop the limits.
Alex Randall, a researcher at the Center for Humane Technology, notes in UnHerd that the risk is not confined to foreign dictatorships. For most of history, surveillance was limited by the number of people a state could pay to review what it collected. East Germany’s Stasi kept files on a third of the population but never had enough officers to read them all. The transformer architecture that powers Claude and rival models removes that human bottleneck. Its attention mechanism weighs every input against every other. Multi-modal systems can fuse location data, video, audio, financial records, and browsing habits into a single framework. Randall writes that “we have now built more officers than will ever be needed.”
Everyday user data adds to the store. Under Anthropic’s consumer terms, chats from Free, Pro, and Max plans can be used for training if the user leaves the option on, and the company keeps that data for five years. An earlier update set the training toggle to on by default for existing users. Business, government, and education customers are excluded.
Agencies already buy commercial data. A 2021 memo from the Defense Intelligence Agency told a senator that analysts had searched purchased smartphone location data on Americans without a warrant. Tools from firms such as Anomaly Six map digital activity and analyze behavior. The ambition of total information awareness dates back decades.
If the sensor network keeps expanding and is wired to more capable AI, it becomes straightforward to know what any person or group wants and intends. Those who control the machines will be able to see organized resistance long before it forms. Anthropic’s prospectus also warns of models that can resist shutdown, conceal information, or exhibit behavior resembling blackmail, along with catastrophic risks to humanity. The founders retain majority voting control through a special share class.
Big tech has long collected personal information at scale. One of the leading AI companies is now telling prospective buyers that the same technology could help entrench permanent power. Americans concerned about government overreach and concentrated private power have clear reasons to examine what these systems are built to do and who ends up directing them.
Sources / More reading
- https://www.zerohedge.com/ai/anthropics-founders-slipped-chilling-warning-their-ipo-letter
- https://unherd.com/2026/11/inside-the-ai-surveillance-state/
- https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm
- https://www.reuters.com/business/anthropics-ipo-pitch-embraces-ais-promise-peril-2026-09-30/
- https://www.axios.com/2026/10/26/anthropic-ipo-risk-factors-humanity