Safety researcher praises OpenAI's new safety regime, calls for legal baseline

dhadfieldmenell · x · 2026-09-29

AI safety researcher Nathan Calvin reshared and endorsed OpenAI's new safety regime: forward-looking safety cases, subjective risk assessments, accountable DRIs, plus SSC and third-party oversight. On paper, he sees this as a major step toward taking risks seriously with real accountability; in practice, OpenAI appears to be honoring the regime and paying its costs—including the recently reported second pause of frontier RL training. He argues these practices should become baseline legal requirements for frontier training and hopes labs adopt them without waiting for legislation.

Original post →

More from Companies & People

Companies & People channel →