More AI safety disclosure sounds good, until the warning itself helps create new risks.

#AISafety #AITransparency #AI风险 https://spaisee.com/article/openai-s-misalignment-framework-tests-how-transparent-ai-safety-can-be