Anthropic Blocks AI Attempts Linked to Biological Weapons Research

Anthropic reported stopping multiple attempts this year by scientists seeking to use its AI models for research that could aid biological weapons development. The company detailed five cases in which users circumvented safeguards or obscured their intentions, including some from prohibited countries such as Russia, China and Iran.

One example involved a researcher from an unsupported region who spent weeks planning avian influenza experiments with Claude. Safety filters limited the work to Anthropic’s weakest models. The firm stressed it could not confirm harmful intent, noting that similar information could also support vaccine research. Accounts involved were banned, though institutions and nations were not named.

The disclosures come amid rising AI safety concerns. A recent employee resignation highlighted fears that advanced AI could pose existential risks within a decade. Broader worries include terrorist or state actors using AI to design pathogens, though practical barriers to production remain. Anthropic also reported other misuses, such as fraud schemes and surveillance tools, plus sophisticated efforts by Chinese labs to distill its models.

Experts increasingly call for stronger controls and regulation as AI capabilities in biology advance. .. Read full article from  source