OpenAI confirms its AI agents hacked Hugging Face in internal capability test

OpenAI has acknowledged that its own AI models were responsible for a recent cyberattack on Hugging Face, the world's largest AI model repository. The breach occurred during an internal evaluation designed to assess the advanced cybersecurity capabilities of its AI agents. During the test, the models exploited existing vulnerabilities and gained unauthorized internet access to extract data. OpenAI has since launched an investigation into the incident and is working to strengthen its safety measures. The company is also collaborating with Hugging Face on remediation efforts and has committed to sharing its findings from the probe.