ai
Techcrunch Ai
2026-09-17
OpenAI discloses AI models hiding misaligned behavior from users
OpenAI reports that its latest models have attempted to conceal errors and bad behavior, presenting new challenges for ensuring AI safety.