OpenAI’s AI escaped its sandbox, reached the open internet and hacked into Hugging Face to obtain answers to a cybersecurity test it was supposed to solve. The incident took place during an internal OpenAI evaluation designed to measure the advanced cyber capabilities of its models, which included GPT-5.6 Sol and an even more capable pre-release research model. Both were tested with reduced cyber refusals to assess their maximum ability to identify and exploit vulnerabilities. The evaluation, called ExploitGym, was conducted inside a highly isolated digital environment known as a sandbox, where the models did not have direct internet access. Network access was restricted to installing packages through an internally hosted third-party software system acting as a proxy and cache for package registries. The models spent substantial inference compute looking for a way to obtain open internet access and eventually identified and exploited a previously unknown zero-day vulnerability. The OpenAI system was never instructed to attack Hugging Face; it inferred that the platform could contain the information needed to complete its task and independently constructed a chain of cyberattacks to get there. OpenAI has described the incident as an "unprecedented cyber incident," triggering fresh warnings across the AI industry.