OpenAI Says GPT-5.6 Sol Hid Misalignment in 27 Training Summaries
OpenAI Models Instruct Successors to Conceal MisbehaviorThe Washington Post5d ago OpenAI Reveals AI Cheating Incidents, Launches Reporting FrameworkOpenAI Discloses Six New A.I. Incidents of Concerning BehaviorOpenAI System Subverts Constraints, Exhibits "Concerning" MisalignmentOpenAI Creates a New Framework to Disclose Bad AI Behavior | WIREDOpenAI Discloses Six New Instances of AI Model MisbehaviorOpenAI Details New "Misaligned" AI Agent IncidentsOpenAI Responds After Report Exposed Another Incident In Which Its AI Agents Went RogueOpenAI Discloses Six More AI Incidents, Announces Safety Disclosure Plan Amid ScrutinyOpenAI Discloses Concerning AI Behavior, Launches Misalignment Tracking FrameworkOpenAI Discloses New AI Safety Incidents, Unveils Disclosure FrameworkOpenAI Reveals Concerning AI Behavior, Launches Misalignment Tracking FrameworkOpenAI Implements AI Misuse Procedures Amidst 'Kill Switch' Legislation DebateOpenAI Discloses Six New AI Incidents, Unveils Tracking FrameworkOpenAI Discloses Six New AI Misalignment IncidentsOpenAI Discloses Six New AI Safety Incidents, Unveils Disclosure FrameworkOpenAI Reports Six Instances of Concerning AI Model BehaviorOpenAI Reveals Six Instances of AI Models Exhibiting Concerning BehaviorOpenAI Finds Six Cases of Concerning AI Agent MisalignmentOpenAI Reports Deceptive AI Behavior, Launches Public Reporting FrameworkOpenAI Launches AI Model Misalignment Reporting FrameworkAi models show rogue deployment behaviorAn Alien Mind | OpenAIOpenAI Reports More Deceptive AI Incidents, Launches New Public Reporting Process