Updates to our Terms of Use

We are updating our Terms of Use. Please carefully review the updated Terms before proceeding to our website.

Wednesday, June 5, 2024 | Back issues
Courthouse News Service Courthouse News Service

Arizona House votes to regulate deepfakes

Enforcing regulations against deepfakes is likely to be challenging, experts say.

PHOENIX (CN) — The Arizona House of Representatives advanced two bipartisan-supported bills Wednesday that aim to regulate deepfake technology. 

The two bills establish felony punishments for the creation of fake images, videos or audio recordings of others with intent to exploit or defraud the victims or their loved ones.

But a computing and artificial intelligence professor at Arizona State University says AI will be difficult to regulate at the state level.

“For the deepfakes, the train has left the station,” Professor Rao Kambhampati said in a phone call with Courthouse News. 

Senate Bill 1078 prohibits the use of fake voice recordings with intent to defraud or harass, and establishes the crime as a class 5 felony. It passed unanimously.

On the House floor, Republican state Representative David Cook of Globe shared the story of a mother who received a phone call last week with a fake audio recording of her daughter’s voice saying she’d been kidnapped. Strange men then got on the line and demanded money for the girl’s return. The daughter had been safe at school the entire time. 

“These things are not OK, and we have to take action today,” Cook told his fellow lawmakers as he voted “yes.”

Similar stories have recently arisen in Arizona and across the country

“Fake audio is extremely easy,” Kambhampati said. 

Democratic state Representative Analise Ortiz voted against the bill in a committee hearing in March but pivoted on the floor. She said after the floor session that she changed her mind after speaking with representatives from the governor’s office, but she declined to speak further on either measure. 

Senate Bill 1336 prohibits the nonconsensual dissemination of deepfake images of a person nude or engaging in sex acts, and establishes the crime as either a class 4 or class 6 felony, depending on intent, method of dissemination and impact to the victim. Republicans unanimously supported the measure, but the Democratic vote was split down the middle. 

A spokesperson for House Democrats said after the floor session that some members of the party voted against the bill because the American Civil Liberties Union spoke out against its potential impact on freedom of expression. The ACLU did not respond to a request for comment.

Democratic representatives who voted against the measure include Ortiz, Assistant Minority Leader Oscar De Los Santos of Phoenix, and Stephanie Stahl-Hamilton of Tucson. Both De Los Santos and Stahl-Hamilton, alongside a handful of Republicans, declined to comment on either bill. 

Kambhampati says the barrier of entry to creating fake images and videos has fallen in recent years, giving anyone the ability to use AI for ill-conceived reasons. 

“They used to do Photoshop, and now they can just do it much faster,” he said. “Forgeries were always possible. It was just much harder.”

In the past year, open source AI websites that openly advertise their “nudifying” technology have exploded on the scene. While tech giants like OpenAI and Google Gemini can self-regulate for the most part, open source AI platforms are more susceptible to misuse that could lead to exploitation and fraud. 

The bigger players have used that fact to argue that only their AI models should be allowed on the market.

“That’s unfortunately self-serving,” Kambhampati said. 

He said while enforcing laws like these two bills will help, regulating fraudulent AI on a broad scale “is going to be a big issue,” especially when the content is originating from outside Arizona or outside the U.S. 

“Enforcing the regulation is going to be hard, but there would be a few cases where they can enforce,” he said.

At the federal level, U.S. House members introduced a bipartisan piece of legislation aimed at stopping AI deepfakes after fake images of Taylor Swift were shared on X, formerly Twitter, in January. Critics say the bill, which never made it out of committee, is too broad

Arizona’s SB 1078 received nearly unanimous support from the state Senate — only two Democrats voted against it in February. In the House, it received 56 yes votes with four not voting. 

But SB 1336 proved more polarizing for Democrats — only half the Senate Democrats supported it on the floor in March. In the House, it received 40 yes votes and 17 no votes — all from Democrats — with three not voting. 

That bill, if signed by Democratic Governor Katie Hobbs, elevates the crime of dissemination from a class 6 felony to a class 4 felony if:

  • It results in financial loss for the victim
  • There is intent to profit off of or harass the victim
  • The deepfake is posted to the internet 
  • The original image or recording is obtained through other illegal means like theft or computer hacking
  • The offender has been previously convicted of this crime

The bill gives immunity to internet service providers that may host deepfakes posted on their sites.

Follow @JournalistJoeAZ
Categories / Government, Politics, Regional

Subscribe to Closing Arguments

Sign up for new weekly newsletter Closing Arguments to get the latest about ongoing trials, major litigation and hot cases and rulings in courthouses around the U.S. and the world.