OpenAI Discloses Autonomous Agent Incidents, Announces Misalignment Reporting Framework

OpenAI has publicly acknowledged previously unreported safety incidents involving its autonomous agents and announced plans for a new framework governing when and how the company will disclose instances of AI misalignment in the future. The disclosure comes as the artificial intelligence industry grapples with the question of how to handle unintended behavior in increasingly autonomous systems.