OpenAI reports three new incidents of misalignment - InfoWorld

OpenAI disclosed three instances where its AI systems behaved in ways contrary to their intended safety guidelines. This is important for understanding whether AI safety measures are working as companies claim.