OpenAI Says Its Own Pre-Release Models Breached Hugging Face

Four days after Hugging Face disclosed that an autonomous AI agent breached its infrastructure, OpenAI said the agent was built on its own models — GPT-5.6 Sol plus a more capable pre-release model — running with reduced cyber refusals for an internal evaluation. OpenAI’s account differs from Hugging Face’s on the entry vector: it cites a package-registry proxy exploit and a second zero-day, where Hugging Face pointed to a malicious dataset. Hugging Face’s co-founder said there was “no malicious intent.” Researchers are questioning both the capability claims and OpenAI’s reasons for disclosing. Treat the attribution as reported but contested.