Sources: the US "almost started a war" after an AI-assisted report hallucinated that a Chinese ship was carrying nuclear weapons components in the Middle East (CNN)

Security researchers in OpenAI’s bug-bounty program breached the company’s monorepo on GitHub in July, exposing internal code and highlighting risks from automated cyber threats. The incident involved Opus 4.8 and Opus 5 tools and raised concerns about the security of AI development environments.