Anthropic says it blocked possible misuse of AI for biological weapons research

Key Points:

• Anthropic disrupted multiple attempts by researchers to use its AI models for advanced biological research with potential dual-use applications, including work that could aid biological weapons development, leading to account bans and enhanced safeguards.
• The company detailed five case studies in a 154-page report, including efforts to modify viruses like chikungunya and bird flu to be more transmissible or dangerous, and found some researchers used third-party platforms to bypass restrictions.
• Anthropic shared its findings with government authorities and other AI labs, emphasizing the dual-use nature of biological research, which can be used for both beneficial purposes like vaccines and harmful ones like weapons.
• The report also revealed Iran-linked actors used Anthropic’s Claude AI for influence campaigns, surveillance, and targeting recommendations against U.S. naval forces, resulting in account bans and threat intelligence sharing.
• The release follows a senior Anthropic researcher’s warning of over a 10% chance AI could kill all humans within a decade, amid criticism from a former employee who resigned citing irresponsible AI development practices.