Anthropic says AI models were used in state surveillance campaigns targeting dissidents

Anthropic says it has disrupted several state-sponsored surveillance operations that used its artificial intelligence models to target dissidents and minority communities. The company said the campaigns were detected and stopped between January and July, and that they originated in China, Iran and West Africa. According to the disclosure, the operations focused on diaspora communities, including Hong Kong pro-democracy activists, Tibetan and Falun Gong communities, and Iranian opponents living abroad.

Sponsored

The company said one Iranian-linked case involved actors developing a method to identify people through their social media accounts. In another case, a contractor working for Malian national security authorities used the model to design software that enabled intelligence gathering. Anthropic said the incidents showed that AI is increasingly being used to support surveillance work that would previously have required a larger engineering workforce.

The report adds to growing concern over how generative AI tools can be adapted for state security purposes. It suggests that the same systems used for ordinary commercial tasks can also be turned toward identifying, tracking and profiling people across borders. That raises particular concern for exiled activists and minority groups who already face monitoring by governments or state-aligned actors.

Anthropic also said it disrupted other misuse attempts involving weapons design, questionable biological research and dating scams. Separately, it accused Chinese developers of using its model to generate responses for users while also distilling that output to improve their own systems. Distillation is a process in which one AI model is used to train another, and the company said some developers had used it without permission.

Sponsored

The disclosure comes amid wider debate over the security risks of advanced AI systems and the difficulty of preventing misuse once models are widely available. It also highlights the cross-border nature of surveillance campaigns, which in this case were said to involve actors in multiple regions and targets spread across Asia and the Iranian diaspora. The report does not say whether any individuals were directly harmed, but it indicates that the operations were active enough to be identified and interrupted over a six-month period.

It remains unclear how many people were affected, what specific agencies or groups were behind each campaign, and whether any of the surveillance efforts continued after the disruptions. Anthropic said the cases were found between January and July, but it did not provide a full breakdown of each operation in the material supplied. Further details may emerge as the company and other security researchers continue to examine how AI tools are being misused in state-linked surveillance.

#Anthropic #artificialintelligence #surveillance #dissidents #China #Iran #WestAfrica