Sam Altman: Pausing frontier RL training to meet alignment and security standards

sama · x · 2026-08-19

OpenAI CEO Sam Altman announced the pause of some frontier reinforcement learning (RL) training to ensure alignment, security, and monitoring standards match the new level of capabilities. He noted that model progress is extremely rapid and that they would act if capabilities outstripped safety. While believing in industry coordination, OpenAI will act unilaterally in the meantime. Altman remains optimistic about their alignment work and committed to widely deploying frontier capabilities.

Related event: OpenAI halts frontier RL training for the first time over safety risks(24 posts)→

Original post →

More from Companies & People

Companies & People channel →