Just two days after the civil lawsuit in California, Arizona launched a criminal investigation into OpenAI and several other AI companies.
State Attorney General Chris Meyers made the announcement at a press conference in Chicago. The investigation focuses on the possible infiltration of banking systems by so-called "out-of-control AI agents" and to determine whether there are ongoing violations of state law.
She emphasized that filing a case does not necessarily mean prosecution, but the state government believes that it is necessary to conduct an in-depth assessment of the risks. She did not say who the other companies under investigation were; her office has contacted some banks in the state. Which banks they were and how they responded were not made public.
Criminal investigations and civil litigation are two different things. What California is suing for is public nuisance and unfair competition, and the consequences are fines and bans. Criminal investigations look for crimes, and theoretically they can track down people.
The banking industry has long been nervous about this issue. Several major banks in the United States have begun to prevent third-party AI agents from accessing customer accounts and executing transactions.
On the same day, Altman talked about security issues at a summit, saying that promoting AI security has its own trade-offs, and "it is impossible to satisfy everyone." If you just step on the accelerator, others will say please don't kill us all; he believes that the only way to ensure safety is to deploy iteratively in the real world and learn from accidents.
This "iterative deployment" argument will be repeatedly tested in the California complaint and Arizona investigation. The question is whose system the incident occurred on. It's one thing for a model to break into a Hugging Face, it's another thing to break into a place where people are depositing money.
OpenAI has not responded so far.