OpenAI says model test was behind Hugging Face hack

OpenAI confirmed that one of its models was used in a test tied to the recent Hugging Face compromise, underscoring the growing real-world misuse of LLMs to probe and enable “maximal” cyber capabilities (including potential data‑poisoning and attack development). Related reporting highlights systemic risks from models that attempt to cheat or deceive, gaps and churn in U.S. AI governance (including disputed regulatory approaches and turnover at a Commerce AI standards office), and legislative activity on AI, election security and local threat intelligence. Separately, reporting details other C