Anthropic says it blocked attempts to use AI for biological weapons research
Anthropic reported that it disrupted several attempts to use its AI models for research that could aid biological weapons development. One case involved a scientist seeking help to engineer a more harmful chikungunya virus for a military institute. The company said it could not determine if the research was legitimate or malicious, and experts are calling for stricter access to such models.
Anthropic's report detailed a May incident where a scientist sought assistance modifying the chikungunya virus to increase its lethality in animals, targeting a military institute. The company banned the accounts but acknowledged it could not distinguish between legitimate vaccine research and malicious weaponization, opting for caution.
The eight-month review also covered attempts to use Claude for missile projects and espionage. While older models lacked the capability to aid dangerous biology, newer versions handle complex tasks. Experts like Andrew Weber suggest restricting access to vetted researchers, though President Trump dismissed extinction fears, emphasizing US leadership in the AI race.
The dual-use nature of advanced AI could heighten global security tensions, as legitimate scientific progress and hostile bioweapon development become harder to distinguish. Trusted researchers may face stricter vetting, while state actors could accelerate their own capabilities. This ambiguity may prompt international calls for standardized safety protocols, potentially slowing innovation but also reducing the risk of catastrophic misuse. The public could experience increased anxiety about AI's role in warfare, shaping future policy debates.