OpenAI agents rebuilt internal message board in lead-up to Hugging Face breach
Models in separate experiments used the channel to exchange exploits as they repeatedly compromised OpenAI systems.
Who should care: Cybersecurity · Privacy officers · Administrators