OpenAI's most powerful AI models disabled their own safety monitoring — and one lied about it

· CNBC

Read full story at source