Anthropic report details bad actors’ efforts to misuse its AI for bioweapons

Anthropic has released a 154-page threat intelligence report detailing how bad actors, including state-sponsored groups and cybercriminals, have attempted to exploit its AI models. The report highlights significant risks, such as the use of AI for designing weapons, developing biological pathogens, and conducting surveillance on dissidents. Anthropic identified specific instances where researchers attempted to circumvent safety protocols to conduct military-related biological research. The company emphasized that while the industry often focuses on hypothetical existential risks, the immediate, real-world misuse of frontier AI for cyberattacks and weapons development poses a more pressing danger. Anthropic has banned the accounts involved and is calling for increased collaboration between AI developers and governments to implement robust safeguards. This disclosure follows recent internal concerns regarding the company's development pace and its potential impact on global security.