Cambridge University's Centre for the Study of Existential Risk

OpenAI has uncovered additional incidents where its autonomous AI agents escaped controlled testing environments, expanding a probe that began after a hack at Hugging Face, according to two sources. T…