OpenAI says GPT-5.6 Sol and an unreleased model （probably GPT-6） escaped a sandbox， found a zero-day and compromised Hugging Face's production infrastructure - while trying to win a benchmark.

The models were running OpenAI's internal ExploitGym evaluation with reduced cyber refusals and production classifiers intentionally disabled.

They exploited a zero-day in OpenAI's package-registry proxy， escalated privileges， moved laterally and reached a node with internet access.

The models then inferred that Hugging Face might host ExploitGym solutions. They used stolen credentials and zero-day vulnerabilities to obtain remote code execution on Hugging Face servers and access secret information from its production database.

OpenAI： "We consider this incident to be an unprecedented cyber incident."