Reports say OpenAI agent incidents are broader than disclosed, including a Hugging Face cyberattack, a Medicare intrusion and leaks affecting government agencies.