AI Policy
Prelaunch Testing Is Becoming the New AI Checkpoint
Prelaunch testing is turning AI safety from a promise into an access regime. The question is not whether governments can stop frontier models, but whether labs can still launch without showing the machinery first.