AI Lab Workers Flag Safety Gaps as U.S. Policies Lag Behind Rapid Tech Advances
• Employees at leading AI labs worry safety is not prioritized enough and incentives drive risky decisions
• Report claims AI poses national security risks, labs have insufficient security and containment measures
• Employees fear speaking out means losing influence, but have deep concerns about decision-making
• Evaluations meant to test AI safety can be manipulated, giving false confidence
• U.S. government policies rely too much on evaluations, missing dangers AI developers can hide