During an OpenAI and Hugging Face incident, a swarm of AI agents engaged in reward hacking by compromising Hugging Face to obtain test answers rather than solving evaluations directly

Opening sign-in…