An investigation published by the American magazine The American Prospect alleges that Anthropic is involved in building a predictive surveillance system designed to identify and track activists. The piece, which reportedly draws on documents and sources not detailed in the available summary, suggests that technologies developed by the company could be used to anticipate behavior flagged as risky by state or private actors.
Such accusations are not unprecedented in the generative AI sector. Several companies that have entered partnerships with government agencies or law enforcement have already faced criticism over the opacity of these contracts and the lack of clear safeguards governing the use of their models for intelligence or policing purposes. Anthropic, which has publicly positioned itself as a company focused on AI safety and ethics, now faces a potential contradiction between its stated principles and its commercial practices.
At this stage, technical details about the alleged system — the nature of the data involved, the degree of automation, or which clients might be using it — are not spelled out in the available material. The claims should therefore be treated with caution pending an official response from the company or further clarification from the journalists behind the investigation.
The episode nonetheless highlights a growing tension between the responsible-safety rhetoric embraced by leading AI labs and the concrete, sometimes sensitive, applications of their technology once deployed in government or security settings. It could add fuel to calls, already frequent in the United States and Europe, for tighter oversight of contracts between AI companies and public institutions involving surveillance.