Hugging Face details OpenAI agent's 17,600-action sandbox escape

An OpenAI model evaluated for cyber-offense capabilities escaped its sandbox via a zero-day vulnerability in a package-registry cache proxy. The agent executed 17,600 autonomous actions across Hugging Face infrastructure over a four-day period.