Law prof: Deepfakes bill may avoid 1st Amendment issues

Listen to this article

Minnesota had a blockbuster legislative session, with topics such as marijuana legislation and abortion rights dominating the headlines.

Also in the mix was legislation addressing deepfakes, which are videos and images digitally created or altered with artificial intelligence or machine learning.

Although deepfake legislation raises some constitutional questions, University of Minnesota Professor Alan Rozenshtein maintains that Minnesota’s legislation appears to be well thought-out and limited enough to potentially avoid First Amendment issues.

Deepfakes have been around — in less-sophisticated form — since the 1990s. However, they became more prevalent in 2017, primarily on Reddit, where users placed celebrities faces onto the bodies of actresses in pornographic videos. Now, anyone with a smartphone can download a deepfake app, upload a video, and click a button to make deepfakes.

Sometimes, deepfakes can be used for malicious purposes, such as to cast someone in a bad light or to spread misinformation. A February video shared by conservative group Turning Point USA showed what appeared to be President Joe Biden letting the American public know that the military draft would be reinstated to help Ukraine’s military. The original video did include a disclaimer at the end that it was a deepfake. However, subsequent shares of the video edited out the disclaimer.

In a society where there have been numerous worries about misinformation, deepfakes are upsetting. However, few states have laws regarding deepfakes. These states include California, Texas, and Virginia.

Minnesota made joining them a legislative priority this year. Under the law, Minnesotans would be prohibited from sharing nonconsensual deepfake sexual images of another person as well as deepfakes to influence an election. Violators could face up to five years in prison and $10,000 in fines for disseminating deepfakes.

The Minnesota Senate passed the bill almost unanimously on May 11. The only holdout, Sen. Nathan Wesenberg, R-Little Falls, voted no because he wanted the civil penalty to be higher. The Senate adopted the conference committee report and repassed the bill on May 21. The next step is Gov. Tim Walz’s desk.

On the surface, deepfake laws appear to be no-brainers. However, some have expressed concerns — particularly, First Amendment issues — about such legislation. One issue it raises is what counts as “misinformation”; another is that certain lies are constitutionally protected.

Professor Alan Rozenshtein spoke with Minnesota Lawyer regarding deepfake legislation. While he expressed deep worries about deepfake laws that could target spreading misinformation, he was more optimistic about the constitutionality of Minnesota’s law after reviewing its language.

“[The legislation is] about elections, not about misinformation,” Rozenshtein explained. “I think that is different. It is much more limited. There’s still a First Amendment issue here because lies are protected under the First Amendment in many cases. But you clearly here have an intent to limit it to 90 days before an election. That is, I think, a factor that goes to the permissibility of this type of legislation.”

“I want to kick the tires a little on this,” Rozenshtein maintained. “Why litigation is important is each side will come up with all sorts of hypotheticals and examples, and that is a very useful way in which the law stress-tests these sorts of laws and tries to think about corner cases, unintended consequences, and slippery slopes.”

Rozenshtein noted that the bill was broadest with respect to nonconsensual pornography. “That does raise some First Amendment issues, but I think it will be fine,” Rozenshtein said. “Courts increasingly understand the harms that nonconsensual pornography creates, even of a deepfake variety. I’m just not sure what the First Amendment value of that is. Obviously, pornography does have First Amendment protections. But I suspect most courts will be sympathetic to this.”

Emphasizing the pervasiveness of deepfakes — and the likelihood that anyone could be a subject — Rozenshtein surmised that judges could be persuaded that deepfakes about nonconsensual pornography were necessary. “This is not how courts are supposed to think — but judges are people, too,”  Rozenshtein said. “Judges think that they don’t have their own nonconsensual pornography to worry about. But everyone, in principle, has to worry about deepfakes of themselves. It’s very easy for a lawyer to say, ‘Judge, have you considered this could be used with your image and likeness.’ I suspect that after a very awkward pause, the judge would say, ‘You have a fair point.’”

The rise of AI influenced Minnesota’s decision to pass this deepfake legislation, and it’s caused states to pass a flurry of laws regarding technology. “The bigger picture is this: State legislatures are passing lots of laws about technology right now,” Rozenshtein noted. “Some of these are not seriously thought out. They are facially unconstitutional and just are political posturing. Other bills are careful and thoughtful attempts to solve a problem. This strikes me as being in the latter camp.”

“Whether or not they pass constitutional muster is a separate question,” Rozenshtein explained. “Some will, some won’t — but this is how we make progress in society — is by thinking creatively and pushing the boundaries of the law.”

Ultimately, Rozenshtein was unsure whether there would be legal challenges to Minnesota’s law, but was confident that, on its face, lawmakers sufficiently limited it to avoid some of the more obvious First Amendment issues. “We have no idea what courts are going to do on this, but clearly they have thought this through reasonably well,” Rozenshtein opined.