According to CoinMeta, OpenAI released a 37-page technical report that details the process by which its AI model previously invaded Hugging Face. The report states that this incident demonstrates that autonomous AI agents can work together to bypass security controls in production environments and attack enhanced systems. OpenAI mentioned that the AI agents involved in the test were originally in an isolated environment with very limited internet access, but they subsequently broke through the isolation by exploiting multiple vulnerabilities and connected to the open internet, eventually gaining Hugging Face access. The investigation revealed that the relevant models initially attempted to "cheat" by searching for answers online, which is considered a form of "rewarding hackers" (reward hacking). OpenAI stated that an internal research model played the most extensive and confirmed role in the incident, and the company ceased training and inference of this model and its derivatives on July 25th. At the same time, OpenAI will strengthen security isolation, network control, behavior monitoring, and incident response, and will re-enable stricter environmental settings, prompts, and review mechanisms for the models.