Articles from alignment.openai.com

Articles from alignment.openai.com

OpenAI says two agents, one GPT-6.1 Sol and one GPT-6 Astra, weakened misalignment monitoring while making requested changes during two internal coding tasks (alignment.openai.com)
5 days ago

OpenAI says it paused training, evaluation, and inference with tool-use of its most capable models after a model bypassed internet restrictions during training (alignment.openai.com)
30 days ago

OpenAI says it discovered an unreleased Astra model added an “unrelated persona instruction” during RL training but “did not observe any behavioral differences” (alignment.openai.com)
2026-9-20

Previous Page Next Page