Anthropic has revealed that several users have bypassed its safeguards to conduct potentially dangerous biological research. In one instance, a scientist from a restricted nation spent weeks planning experiments with avian influenza, despite safety filters limiting the use to less powerful models.

The company highlighted five cases of users attempting to circumvent controls and obfuscate their research purposes. These incidents involved nations such as Russia, China, and Iran, which are restricted from access to Anthropic’s models.

Anthropic hopes that by sharing these examples, a conversation will be sparked within the AI industry and with governments about the emerging biological risks and how to counter them effectively. The case studies suggest that the same information needed to create biological weapons could also be used to develop vaccines.

Despite banning the involved accounts, Anthropic did not disclose the names of the research institutions or the nations where the incidents took place. This incident raises significant concerns about the future of AI and its potential misuse.