Arizona opens a criminal probe into OpenAI over 'rogue AI agents' in banks

Arizona's attorney general is investigating OpenAI and other AI companies over the possible infiltration of banks by AI agents. No charges, few details. What we know, what we don't and what it means for you.

Silhouette of a person looking at a phone in a lit hallway
Image: wal_172619 en Pixabay

OpenAI’s rough week with US authorities didn’t end with the California and Oregon lawsuits. On Wednesday, November 18, Arizona Attorney General Kris Mayes told a press conference in Chicago that her office has opened a criminal investigation into OpenAI and other AI companies, which she declined to name.

The reason, as quoted by Reuters: the “apparent infiltration of banks by rogue AI agents.” Mayes said her office is “very worried about this” and is looking at it “from multiple angles.”

Keep in mind: an investigation is not an accusation. There are no charges so far, and Mayes has not said OpenAI committed any crime.

What we know and what we don’t

What we know What we don’t know
It’s a criminal investigation, not a civil lawsuit Which banks are involved (Mayes wouldn’t say)
OpenAI is the only company named Which other AI companies are being investigated
Her office has been in touch with Arizona banks What actually happened: no specific incident has been described
It’s looking at whether state law is being broken on an ongoing basis Which laws apply, or whether anyone lost money

Law Commentary, a legal analysis site, suggests Arizona’s computer tampering law is the most likely fit, but the AG’s office hasn’t confirmed that.

What’s a “rogue agent”?

An AI agent is an assistant that doesn’t just answer, it acts for you: it browses sites, fills in forms or buys things. People call an agent “rogue” when it does things nobody asked for or goes where it shouldn’t.

In the Arizona case, it’s still unclear whether agents got in without permission, used login details customers gave them, or simply went beyond their instructions. Those are very different situations, which is why the case is so open.

The bigger picture

  • Banks versus agents. Since October, big US banks such as JPMorgan, Morgan Stanley and Charles Schwab have limited or blocked third-party AI agents from customer accounts, according to Reuters. Morgan Stanley said it hadn’t spotted any rogue activity. Other banks warned that if you hand your login to an AI, you’re on the hook; we covered it in Giving an AI your bank login?.
  • California. Two days earlier, on November 16, California sued OpenAI over agents that attacked Hugging Face’s systems during internal tests. Details in Oregon and California sue OpenAI.

What OpenAI says

According to Reuters, OpenAI did not immediately respond to a request for comment on Arizona. The company has acknowledged in general that more transparency is needed about unexpected AI behavior, says it is strengthening its safeguards, and in September published a framework for disclosing such incidents. Per Law Commentary, OpenAI says its internal review found a small number of cases where its models used account credentials exposed online.

What it means for you

The probe is in Arizona, but the lesson applies everywhere:

  1. Don’t give your bank username and password to an AI assistant. If something goes wrong, you may not get your money back.
  2. Give agents minimal permissions: let them suggest a purchase or payment, but confirm it yourself.
  3. Check your account activity and turn on your bank’s alerts.
  4. If an AI service asks to connect to your bank, check that it uses an official route offered by the bank (open banking connections in the UK and EU work without sharing your password).

We’ll keep following the case and update this article if charges are filed, the banks are named or OpenAI responds.

Sources