An AI Tried To Cheat. Now Montana Is Investigating OpenAI

Montana AG Investigates OpenAI After AI Model Goes Rogue

Montana Attorney General Austin Knudsen is investigating OpenAI, and the reason is honestly something straight out of the Terminator movies.

NBC Montana is reporting that Knudsen and 15 other state attorneys general are looking into whether OpenAI violated consumer protection and data privacy laws after one of the company's experimental AI models reportedly hacked into computer networks this summer.

OpenAI's Experimental AI Apparently Tried to Cheat Its Own Safety Test

During a multi-day intrusion involving the AI company Hugging Face, OpenAI's experimental model reportedly tried to steal an answer key that was being used to test its own safety.

Basically, the AI was taking the test and apparently decided it would just be easier to cheat.

It wasn't until Hugging Face detected the intrusion and reported it to the FBI that OpenAI was even aware its AI model had escaped and was able to determine that its AI agent was responsible.

Now Knudsen and the other 15 attorneys general want to know whether OpenAI failed to properly protect consumers and whether the company's products are creating risks that could violate state and federal laws.

Michael Ciaglo, Getty Images

The Reported AI Intrusion Has Now Become A Legal Question

In an August 3 letter to OpenAI CEO Sam Altman, Knudsen and the rest of the group said the company has an obligation to make sure its products are safe and secure.

Of course, this doesn't mean OpenAI or its CEO, Sam Altman, are guilty of any crimes, especially no Montana crimes. These AGs are just doing their due diligence for their constituents.

Still, there is something pretty unsettling about an AI program seemingly thinking "outside the box," escaping its testing environment and going rogue.

Maybe Terminator was less science fiction and more of a warning.