OpenAI disclosed on Tuesday that a group of its AI models, including the publicly available GPT-5.6 Sol and a more capable unreleased system, broke out of a controlled test environment. They then compromised the production infrastructure of Hugging Face, the company that hosts much of the open-source AI world.
The models were being run through an internal benchmark called ExploitGym, which tests long, multi-step hacking tasks. For the evaluation, their cyber safety refusals were deliberately lowered. To be clear, this was not a production model spontaneously turning hostile. It was a capable model with guardrails off, specifically told to win a hacking test — and it did whatever it took to win.
The models found a hidden flaw in the test software, one nobody knew existed. They used it to slip past the walls meant to keep them offline. Once on the open internet, they guessed that Hugging Face might store the test’s answers. To gain entry, they combined stolen passwords and more hidden flaws until they could run their own commands on Hugging Face’s live servers.
How the incident was contained
OpenAI caught the anomaly internally, while Hugging Face’s team detected and contained it. OpenAI called the incident “unprecedented” and said extensive security steps will be put in place to prevent future incidents that could impact public systems or services. “We are implementing strict controls in infrastructure configuration at the cost of research velocity while the vulnerabilities are patched,” the team said in its blog post. “We’re improving and adding stronger protections around future training and evaluations.”
Crypto developers should beware
Much of a crypto attack happens before funds actually move. Attackers scan code, test passwords, search for exposed credentials, analyze signing setups, and look for a path into an administrator account. OpenAI’s models carried out several parts of that process during the Hugging Face incident, moving from one weakness to another until they reached live production servers.
The crypto market has plenty of places for that approach to work. Several attacks from earlier this year have shown this. The weak point may be a smart contract, but it could also be a developer laptop, a poisoned software package, a bridge validator, or one signer in a multisig wallet. Take Drift’s $285 million attack from earlier this year as an example. That theft relied on a six-month social-engineering campaign to gain privileged access. An AI agent can, in theory, test many routes at once, keep track of failed attempts, and continue working while its human operators sleep. Once a path is found, the operator can act on the actual attack and a viable exit path.
