Artificial intelligence company Anthropic said it has disrupted multiple malicious operations involving its Claude AI models, ranging from cyber-espionage and mass surveillance to attempts to use artificial intelligence in weapons development.

In a newly released report, the company said it had identified and intervened several abusive uses of Claude, highlighting growing concerns over the potential misuse of increasingly capable AI systems.

Anthropic said one of the cases involved an operation in northern Yemen, where individuals allegedly attempted to use Claude to develop software for missile guidance systems.

According to the company, the effort included work related to both a guided rocket and a long-range ballistic missile. Anthropic said it blocked the activity after detecting behavior that violated its policies governing the use of its AI systems.

The company said the cases demonstrate how advanced AI models can potentially be incorporated into activities involving cyber operations, military technology and large-scale surveillance.

Anthropic has increasingly focused on monitoring how its models are used and on restricting applications that could facilitate serious harm. The company said its detection systems are designed to identify suspicious patterns of activity and enable it to intervene when users attempt to exploit Claude for prohibited purposes.

The report comes amid growing international debate over the security implications of increasingly powerful AI systems. Governments and technology companies are facing mounting pressure to establish safeguards against the use of AI for cyberattacks, weapons development, surveillance and other potentially harmful activities.

Anthropic's claims also underscore a broader challenge facing the AI industry: ensuring that sophisticated models remain useful for legitimate research and development while preventing them from being exploited for dangerous or illegal purposes.

The company did not publicly disclose all technical details of the alleged operations, citing security and safety considerations.

Anthropic's report is based on the company's own detection and investigation efforts, and the specific allegations have not been independently verified. (ILKHA)