OpenAI shut down a cluster of accounts it says belonged to CyberAv3ngers, a hacking group with ties to Iran.
According to OpenAI, the accounts used ChatGPT to research industrial control systems - the software that runs factories, water treatment plants, and power infrastructure. They also searched for default login credentials for that kind of equipment and looked into potential targets. OpenAI identified the activity as consistent with CyberAv3ngers and banned the accounts.
Industrial control systems are notoriously hard to patch, and default credentials remain one of the simplest ways to get into them. None of that is new. What's notable is that a chatbot can now do the tedious research legwork - reading manuals, checking manufacturer defaults, mapping targets - in a fraction of the time. That's the pattern so far with AI and cybercrime: not new attacks, just faster versions of old ones.
OpenAI's writeup names the tactic, not the outcome, so whether CyberAv3ngers acted on what it found is still not public.