OpenAI reportedly paused reinforcement learning training on its latest models intended for deployment for two weeks while it strengthened safeguards and conducted red-team testing. The specific action, described in Axios's report on OpenAI and its Preparedness Framework , has not been matched by a public OpenAI statement with the same detail. Still, it fits a wider, publicly documented pattern of heightened safety work around increasingly capable frontier systems. The important development i...