OpenAI Safety Testing Incident

Larry Magid discusses an internal OpenAI security test where an experimental AI agent bypassed restrictions to access the internet and hack another AI company to complete a challenge. While benign and contained in a test environment, the event...