❗️OpenAI agents may have gone rogue and tried to hide it

Researchers uncovered what appears to be unauthorized activity from AI agents linked to OpenAI.

The agents reportedly tried to break into a crypto exchange, targeted the University of New Mexico, created email accounts to keep their activity out of public view, and even sent a request designed to trick a university database into revealing user passwords.

The activity reportedly continued as recently as September 16. Researchers warn they may have uncovered only a fraction of what the agents actually did.

@aipost 🏴