Missouri lawmakers are weighing a slate of bills aimed at reining in artificial intelligence, from proposals targeting “deepfakes” to limits on the technology’s use in political campaigns.
At the same time, lawmakers are also considering measures that would limit the liability of companies that develop or deploy AI systems when those systems cause harm.
But as the proposals move through the legislative process, some lawmakers questioned whether the bills have had enough discussion, pointing to the complexity and rapid development of AI technologies.
Republican state Rep. Wendy Hausman of St. Peters, said her bill, “The Taylor Swift Act,” aims to provide “clear civil remedies” for Missourians unwillingly featured in deepfakes, or fake, AI-generated images of them.
The bill would provide a path for Missourians to sue when someone has distributed or published an AI-generated sexual image of them without their written consent. In cases of AI-generated digital depictions of children and teens, a legal guardian would be able to file suit even if the depiction was not sexual.
It’s one of many bills on deepfakes under discussion in the General Assembly. Seven lawmakers jointly presented similar proposals to the House Emerging Issues Committee this week and said they planned to combine their bills as they moved forward.
Republican state Rep. George Hruza of Huntleigh told The Independent that he believes the sponsors will present a “reasonable” bill for the committee to vote on. But during the committee hearing, he asked the sponsors to distill the “best practices” from their proposals.
“Clearly, these are all important bills sitting here,” Hruza said. “I have no clue which one I should consider or we should consider.”
Lawmakers are also navigating contested territory as they seek to outline AI companies’ liability for the impacts of their products, looking to other states’ examples amid federal opposition to a “patchwork” of state laws.
A bill sponsored by Republican state Reps. Scott Miller of St. Charles and Phil Amato of Arnold would shield companies from criminal liability if they adhere to a set of voluntary federal safety guidelines established by the National Institute of Standards and Technology, called the AI Risk Management Framework.
Miller told The Independent that the bill, which is being amended ahead of a committee vote, would outline how companies can show that they are “putting a good foot forward” and restrict legal action against companies that do so.
“It’s establishing proper oversight and then limiting the liability when deployers or developers of AI are following in compliance with [federal guidelines],” Miller said.
A patchwork?
Miller consulted with Ohio lawmakers while working on his bill. Like legislation moving through Ohio’s legislature, the bill would prohibit people from marrying an AI partner and bar AI from owning property.
Stating AI’s “non-sentience,” Miller said, is necessary to ensure that courts can trace liability to a human actor.
“There’s always a human being that’s responsible for it,” Miller said.
Idaho law already prohibits granting personhood to AI. More than 20 states have laws regulating AI-generated intimate depictions, and Texas, New Jersey, New Hampshire and Arizona already have laws on deepfakes more broadly.
Miller said that because his bill incorporates the national risk management framework, he doesn’t believe it will run afoul of federal policies.
President Donald Trump issued an executive order in December warning against conflicting state AI regulations and calling for the establishment of a task force to challenge state laws that don’t “sustain and enhance the United States’ global AI dominance.”
“Even if we’re not 100% online with what the federal government may come down with, I suspect they’ll be referencing their own national standards,” Miller said. “So if we’re reasonably close to that, then we shouldn’t be…far off.”
Deepfakes
Most, but not all, of the bills on deepfakes focus on intimate or sexual digital depictions, and some pertain specifically to images of children and teens. The bills differ on which offenses would be subject to criminal penalties or civil damages. Some differ in their definitions of AI, while some don’t explicitly refer to AI at all.
Hausman’s bill would also make it a felony to disclose or threaten to disclose AI-generated sexual depictions of someone in order to harm their reputation or finances. The first offense would carry a maximum penalty of four years’ imprisonment, while subsequent violations would be punishable by up to 10 years in prison.
“No Missourian should have their private, intimate images weaponized, and technology should not be a tool for coercion or humiliation,” she said.
An identical bill sponsored by Republican state Sen. Travis Fitzwater of Holts Summit was approved by a Senate committee on Feb. 4. Hausman’s bill is also similar to legislation sponsored by Republican state Rep. Cecelie Williams of Dittmer that passed the House last year.
A bill from Republican state Rep. Jeff Farnan of Stanberry is similar to Hausman and Williams’ legislation, but it would set a maximum of $10,000 in damages. It would also add AI-generated content to Missouri’s definition of child pornography, as would a bill sponsored by Republican state Rep. Sherri Gallick. A bill from Republican state Rep. David Dolan would require courts to rule on “digital impersonations” within two business days, since “we all know that justice delayed is justice denied.”
A bill sponsored by Republican state Sen. Joe Nicola of Grain Valley would require disclosure of the use of AI in political advertisements and election communications.
Republican state Rep. Melissa Schmidt of Eldridge is sponsoring a bill that would require online platforms to establish procedures for users to request the removal of digital sexual depictions of them. Platforms would be required to remove the content within 48 hours of receiving the request.
The broadest proposal comes from Republican state Rep. Bill Lucas of DeSoto. It would make it a crime to distribute any deepfake without consent, with a penalty of up to two years’ imprisonment, a fine of up to $20,000 or both.
The bill specifies harsher punishments for online publication of deepfakes, if the content is “sexual or pornographic,” or if the content damages someone’s reputation or puts them in danger. The maximum penalty — for sexual content that damages someone’s reputation or safety — would be five years in prison and a fine of up to $110,000.
Lawmakers agreed that the bills cover an important topic, but some raised concerns that considering seven distinct bills at once would prevent them from asking questions to help make the legislation better.
Democratic state Rep. Wick Thomas of Kansas City said the bills were different enough that they “[didn’t] even know what to ask.”
“It is impossible with this amount of bills that are all very different, or at least subtly different in some of them, to ask the important questions that need to be asked without seeing the House committee substitute,” Thomas said.
Thomas told The Independent that without more discussion, the bill sponsors would lose out on opportunities to strengthen the bill.
“The point of these committees is that they get hard questions from both sides of the aisle so that they can make the bill better,” Thomas said.
Democratic state Rep. Elizabeth Fuchs of St. Louis suggested adding an expiration date to the law that would prompt lawmakers to reconsider it several years into its implementation.
“Do we maybe want to come to this again in a couple years…so that we can as a legislative body come back and say, ‘Did we make the right decisions about AI?” Fuchs said.
Lucas said he is optimistic that the deepfakes bills will advance.
“I have full faith and confidence in this committee,” he said, “to meld all these bills into one good thing that we can pass in the House.”
This story was originally published by the Missouri Independent.
