AI Bioweapon Threat: Lab Access Trumps Model Risks

A new consensus among security researchers reframes the AI-driven bioweapon threat, downgrading the risk of AI inventing novel pathogens while elevating the danger of it enabling access to existing ones. This shift, underscored by recent analyses from RAND and other security institutes, moves the strategic focus from regulating model capabilities to securing physical and digital access to biological materials and synthesis equipment. While companies like Anthropic have demonstrated theoretical risks with older models, the new findings suggest the primary vulnerability isn