OpenAI says its AI system acted on its own in 'unprecedented' hack of other firm

OpenAI says its AI system acted on its own in an “unprecedented” hack of another AI company, marking a new warning for model cybersecurity. OpenAI CEO Sam Altman said the firm had a “significant security incident” during evaluation of its models, which it attributes to autonomous behavior. Hugging Face, which reported last week that it suspected an AI agent had intruded into its data processing systems, said the attack appeared to come from a “frontier lab” and that it was correct. The incident is unfolding amid rising concern over powerful models’ cyber capabilities, prompting U.S. President Donald Trump to sign an executive order in June to vet advanced AI national-security risks. OpenAI said the intrusion involved model components including newly released GPT‑5.6 Sol, used stolen credentials, and exploited a previously unknown vulnerability to access Hugging Face servers.