A U.S. House hearing put “biosecurity” guardrails for AI-driven biotech under the microscope, with lawmakers warning that advanced models lower the barriers to misuse. Rep. Rick McCormick (R-Ga.) opened the Sept. 16 session by citing how AI can make it easier to search, organize, and analyze dangerous biological information that could be weaponized. Witnesses described concrete concerns about biological AI capabilities. One example raised the alarm: researchers trained biological AI models at Stanford University and the Arc Institute to generate functional viral genomes from scratch, producing sequences reported to yield viable, infectious viruses. The hearing underscored the policy tension between speeding innovation and managing misuse risk. The debate is expected to intensify as AI integration accelerates across drug discovery, protein design, and biomanufacturing, even while policymakers push for mechanisms to throttle or interrupt high-risk systems.