Sen. Cantwell’s AI audit proposal and the World Health Organization’s questions to Russia may show that an outside check is only as good as the information the checker can get

Sen. Maria Cantwell released an AI safety framework that would bar the most powerful AI models from release until an independent audit confirms they meet federal standards, and would require developers to give testers access. Leaders of top AI companies had earlier signed a voluntary accord that includes audits by an outside firm. The World Health Organization says it still does not know what killed a worker at a Siberian plague lab and needs timely, regular information from Russia; Russian health officials say no plague has been found among her contacts. Read together, they may show that an outside check is only as good as the information the checker can get.

Theory - analysis