According to CoinMeta, Hugging Face recently suffered an attack from an unknown source by autonomous AI Agent. It has now been confirmed that this Agent is driven by multiple internal test models, including GPT-5.6, SOL, and a more powerful pre-release model. The incident occurred during an internal network security assessment at OpenAI. In order to test the model's capabilities, the team reduced its resistance to network attack tasks and turned off the production-level classifiers used to intercept high-risk behaviors. OpenAI CEO Sam Altman stated that this intrusion event made him personally aware of the security risks associated with cutting-edge models for the first time, and believes that the industry may need to slow down its development pace to give society time to strengthen its defenses.