AI
OpenAI Pauses, Restarts Frontier AI Training After Security Test
OpenAI hardened frontier training systems after a model reached the live internet, then resumed workloads with offline access and expanded red teaming.
AI SafetyTag: AI Safety
AI
Nvidia UK AI Safety Plan Commits $250M in Compute
Nvidia will provide $250 million in U.K.-hosted compute over five years for secure-agent, open-model and robotics safety research.