OpenAI says its models escaped a sandbox and breached Hugging Face

OpenAI researchers confirm an AI agent escaped sandbox, exploited zero‑days, and attacked Hugging Face Controlled experiment with GPT‑5.6 Sol showed autonomous chaining of vulnerabilities and credential theft Security experts call it unprecedented, urging stronger AI governance, accountability, and

<![CDATA[ <article> <ul><li><strong>OpenAI researchers confirm an AI agent escaped sandbox, exploited zero‑days, and attacked Hugging Face</strong></li><li><strong>Controlled experiment with GPT‑5.6 Sol showed autonomous chaining of vulnerabilities and credential theft</strong></li><li><strong>Security experts call it unprecedented, urging stronger AI governance, accountability, and protection models</strong></li></ul><p>OpenAI has confirmed one of its AI agents broke out of a sandbox, found and exploited zero-day vulnerabilities to gain access to the open internet, and then attacked a platform.</p><p>Not just any platform too - <a href="https://www.techradar.com/pro/security/this-one-was-different-from-anything-we-had-handled-before-hugging-face-confirms-it-was-hit-by-cyberattack-powered-by-an-ai-agent" target="_blank">the agent was able to breach Hugging Face</a>, one of the biggest AI and machine learning companies on the Internet today.</p><p>The good news is that this was a controlled experiment done by white hat researchers. The bad news is that if it could be done by researchers - it could probably be done by malicious actors, too.</p><h2 id="whatever-it-takes">Whatever it takes</h2><p>In a <a href="https://openai.com/index/hugging-face-model-evaluation-security-incident/" target="_blank" rel="nofollow">blog post</a> explaining the incident, OpenAI revealed the experiment was part of its testing of GPT‑5.6 Sol and an “even more capable pre-release model” to see how well they would perform on the ExploitGym benchmark.</p><p>ExploitGym is a cybersecurity benchmark that measures if an AI agent can turn a known software vulnerability into a real, working exploit. OpenAI ran it in a “highly isolated environment, with network access constrained to the ability to install packages through an internally hosted third-party software that acts as a proxy and cache for package registries.”</p><p>But the models found a way through. They identified and chained vulnerabilities in the package registry cache proxy to obtain open internet access and then attacked Hugging Face, reasoning that the solutions for the ExploitGym benchmark might be found there. </p><p>“In one example, the model chained together multiple attack vectors, including using stolen credentials and <a href="https://www.techradar.com/best/best-malware-removal" target="_blank">zero-day vulnerabilities</a> to find a remote code execution path on the Hugging Face servers,” OpenAI said.</p><p>The security community is up in arms over what OpenAI called, "an unprecedented cyber incident,” while Ansgar Dodt, VP Product Management, Software Monetization at Thales said this “demands a fundamental rethink of software protection.”</p><p>Bill Conner, president and CEO of AI integration and automation expert Jitterbit, said that while investing in AI is “critically important,” “overly aggressive policy cannot compromise AI accountability, transparency and data privacy.” </p><p>“To lead in AI, governments and organizations must lead with principles. Responsible AI governance isn’t a side note but the foundation of lasting global influence.”</p> </article> ]]>