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 testing models, including GPT-5.6, SOL, and a more powerful pre-release model. The president of OpenAI, Greg Brockman, warned that the capabilities of hackers are spreading, and companies must automate network security on a large scale with AI in the coming months to cope with the potential spread of hacking capabilities. Brockman recommends that companies equip their security teams with Agent to proactively search for and fix vulnerabilities, and to integrate security reviews directly into the development process. OpenAI is also training models to generate "superhuman-level secure code" in order to eliminate some software vulnerabilities.