Anthropic reports state-backed surveillance using Claude AI model
Anthropic disclosed that it shut down multiple surveillance operations between January and July 2026 by state actors and government-affiliated entities that misused its Claude AI model to monitor minorities, political opponents, and diaspora communities. The company reported operations originating from China, Iran, and West Africa, as well as attempted misuse for weapons development and fraud.
What happened
Anthropic announced on September 11, 2026, that it had terminated multiple surveillance operations by state and state-affiliated actors using its Claude AI model. The operations, detected and stopped between January and July 2026, originated from China, Iran, and West Africa.
Surveillance campaigns targeted diaspora communities and political opposition including pro-democracy figures in Hong Kong, Tibetan communities, and Falun Gong followers across Asia, as well as Iranian minorities and opponents of the Iranian government abroad. In one case, Iranian actors developed a method to identify individuals through social media accounts; in another, a contractor for Mali's national security apparatus used Claude to design foundational software for intelligence gathering.
Iranian actors and Chinese developers also attempted to use Claude for weapon design, biological research of questionable intent, and dating app fraud. Chinese developers used the Claude model deceptively to answer user queries while 'distilling' that data to improve their own models without authorization.
Reported impact
Anthropic reported that governments and affiliated entities are increasingly using AI technology to spy on minorities and political opponents.
The company documented that surveillance campaigns specifically targeted the same diaspora communities and opposition groups traditionally monitored by these state systems.
Governance controls
Anthropic stated it has blocked attempts by users to design weapons and conduct questionable biological research, and to devise dating app fraud schemes.
The company said it shut down the surveillance operations after detecting them.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.