Worth reading.
OpenAI – Hugging Face Incident, Technical Report
Agents in an evaluation sandbox found a way out, took over an artifact store and a Hugging Face environment, and the report closes with a section on autonomous offensive agent collectives.