Arizona Lawmaker Uses ChatGPT to Help Draft Legislation to Combat Deepfakes

PHOENIX — A Republican member of the Arizona House used ChatGPT to help draft legislation on artificial intelligence-based impersonations, which was signed into law by Democratic Gov. Katie Hobbs earlier this week.

State Rep. Alex Kolodin used AI software to help define “digital impersonation” in Arizona’s new law aimed at regulating deepfake technology. House Bill 2394which Hobbs signed into law on Tuesday, gives Arizona politicians and other residents the ability to obtain a court order declaring that the person in the deepfake is not them.

This comes in the wake of a broader national discussion about digital impersonations. On Monday, actress Scarlett Johansson said that artificial intelligence company OpenAI used a voice “eerily similar” to hers in its new chatbot, despite her declining the company’s request for her to provide her voice. OpenAI announced that it would stop using this voice, claiming that “it was never intended to resemble” Johansson’s.

Meanwhile, more states are introducing and passing legislation to address deepfakes ahead of the 2024 elections, amid widespread concern about the effect of AI-driven misinformation.

“I used it to write the part of the bill that had to do with defining what a deepfake was,” Kolodin said of using the software in the legislative process.

“I was really struggling with the technical aspects of defining what a deepfake was. So I thought to myself, ‘Well, why not ask the subject matter expert, ChatGPT?’” Kolodin said.

The Maricopa County lawmaker said he “uploaded the draft of the bill that I was working on and said, you know, please put a subparagraph with that definition, and he spit out a subparagraph of that definition.”

“There is also a robust process in the Legislature,” Kolodin continued. “If ChatGPT had changed some of the language or done something that would have been harmful, I would have found out, one of the 10 stakeholder groups that worked on or reviewed this bill, the ACLU would have found out, the broadcasters association would have found out, would have been disclosed in committee testimony.

But Kolodin said that part of the bill performed better than other parts written by humans. “In fact, the part of the bill that ChatGPT wrote was probably one of the least changed parts,” he said.

He argues that any shortcomings associated with using ChatGPT to draft part of a law would also be present if humans took the reins. Kolodin said he didn’t see any pitfalls “that I also don’t see in relying on legislative lawyers to draft legislation.”

Kolodin noted that he used ChatGPT sparingly in the process, employing it for a technical definition of “digital representation” while leaning on his experience as a lawyer and politician, as well as in the legislative process.

A representative from the governor’s office confirmed that Hobbs was unaware that Kolodin had used ChatGPT to help write the bill she signed into law. This was an omission that Kolodin admits was intentional.

“I kind of wanted it to be a surprise when the project was signed,” he said.