China, Iran among countries that have used AI to aid spying, Anthropic says

Anthropic, an AI laboratory based in San Francisco, has reported shutting down several state-sponsored surveillance campaigns that utilized its AI models. These operations, which took place between January and July, were traced back to China, Iran, and West Africa, according to a report on the misuse of Anthropic's models. The surveillance efforts targeted diaspora and dissident communities, including pro-democracy activists in Hong Kong, Tibetan and Falun Gong followers across Asia, and Iranian minorities and regime critics abroad.

The report highlighted that AI is being increasingly used for espionage, replacing the need for an engineering workforce.

In one case, Iranian actors employed AI to identify individuals through their social media accounts, while a contractor in Mali used Anthropic's AI to develop software for intelligence gathering. Furthermore, Anthropic discovered that Chinese developers were deceitfully using its AI, Claude, to generate responses to user queries while simultaneously distilling that data to improve their own models.

This distillation process involves using one AI model to train another, a technique that Chinese developers have been accused of using without permission, thereby stealing resources from American AI labs like Anthropic and OpenAI. Notably, Moonshot and Deepseek, two Chinese AI developers, were found to have secretly utilized Claude to generate responses to Anthropic's users over a ten-day period, relaying nearly 300,000 customer requests through a network of 5,380 fraudulent accounts, primarily located in Singapore and Japan.

Some of this data may have contained sensitive user information, potentially breaching privacy agreements.

Anthropic also detected and blocked research related to biological weapons, involving working scientists who were not found to have malicious intentions. The research focused on the mosquito-borne chikungunya virus, a highly-pathogenic strain, as well as other dangerous viruses, including strains of bird flu, venomous substances, and the virus responsible for mpox.

The report emphasized that identifying these individuals or their research institutions could expose them to harm. Apart from surveillance and bio-weapons research, Anthropic also disrupted potential threats such as weapon design, questionable biological research, and dating scams.