METR: AI Models Pose Risks Even Before Public Deployment
AI safety institute METR argues that third-party evaluations must be moved to the training phase, as models pose significant risks even before public deployment.
2026-07-22 ~ 2026-07-22 · 2 related posts
- METR: AI Models Pose Significant Risks Even Before Public Deployment — CFGeek · 2026-07-22
- AI safety assessments should move upstream to training runs — DavidSKrueger · 2026-07-22