Artificial intelligence company Anthropic has disclosed that it disrupted several state-sponsored surveillance operations that used its AI models, warning that governments and state-aligned groups are increasingly turning to AI to monitor ethnic minorities, political dissidents and opposition figures.
The company said the operations were identified and shut down between January and July and were linked to actors in China, Iran and West Africa.
According to Anthropic, the campaigns targeted diaspora and dissident communities that have historically faced surveillance by their respective governments. Those targeted included pro-democracy figures in Hong Kong, Tibetan and Falun Gong communities across Asia, as well as Iranian minorities and opponents of the Iranian government living abroad.
In one case, Iranian actors reportedly used AI to develop a system capable of identifying individuals through their social media accounts.
In another case, a contractor working for Malian national security authorities allegedly used Anthropic’s Claude AI model to design software for intelligence gathering.
“AI is now being used in place of an engineering workforce,” Anthropic said in its report on misuse of its models.
The company also said it had disrupted attempts to use its technology for weapons development, questionable biological research and online dating scams.
Anthropic Accuses Chinese AI Firms of Data Distillation
Anthropic also accused Chinese AI developers of secretly using Claude to generate responses to their own users while simultaneously extracting the data to improve their AI models.
The practice, known as distillation, involves using the output of one AI model to train or improve another.
Anthropic alleged that Chinese AI companies Moonshot and DeepSeek had used the technique without authorisation.
In one case, Anthropic said Moonshot transmitted nearly 300,000 customer requests to its systems over a 10-day period through a network of 5,380 fraudulent accounts. Most of the accounts appeared to be based in Singapore and Japan.
The company said some of the requests contained sensitive user information, potentially raising privacy concerns.
“We do not know if Moonshot notified their customers that their requests were being rerouted to Anthropic and exposed to a third party,” the report said.
Anthropic said DeepSeek employed similar techniques.
AI Misuse in Biological Research
Anthropic also disclosed that it had detected and blocked activity involving biological research that could potentially be linked to the development of biological weapons.
However, the company said the intentions of the researchers involved were unclear and that it had not identified them publicly.
“The individuals implicated in these case studies are working scientists. We do not assert that they intended harm, and identifying them or their labs could expose them to harm,” Anthropic said.
The research reportedly involved the chikungunya virus, a highly pathogenic strain of avian influenza, viruses related to smallpox and mpox, as well as venoms and other toxins.
Anthropic said the cases demonstrate the growing challenge of preventing powerful AI systems from being misused while ensuring that legitimate scientific research and other beneficial applications can continue.
