Artificial Intelligence study committee considers potential recommendations to lawmakers

A Wisconsin study committee met Thursday to discuss regulations of artificial intelligence as well as investments in AI that it will recommend to the state Legislature. 

The committee includes four lawmakers — Sen. Julian Bradley (R-New Berlin), Rep. David Armstrong (R-Rice Lake), Sen. Tim Carpenter (D-Milwaukee) and Rep. Clinton Anderson (D-Beloit) — as well as eight members of the public. Members have been meeting monthly since July to hear from an array of stakeholders about the ways that AI is currently being used in the state, issues that have or could arise from the technology and ways to address those and how the state could capitalize on potential benefits.

The study committee is the latest effort as Wisconsin government and business leaders attempt to confront increased use of and ongoing advancements of AI technology. Gov. Tony Evers and Assembly Speaker Robin Vos both organized task forces in the last year to address the issue.

Bradley, the committee’s co-chair, noted at the end of the meeting that many of the issues discussed will need continued discussion from lawmakers and members of the public in the future as AI continues to develop and become a bigger part of Wisconsinites’ lives. The committee is considering recommending an ongoing and more permanent body that will continue looking into AI after members end their work this year.

“I believe that a lot of these start with an ongoing committee. There’s a lot of things that say we need to continue to explore… because the technology is changing,” Bradley said. “Things are changing constantly, so we want to be able to have public members with the Legislature come together and say, ‘Hey, maybe not yet,’ or OK, it’s time now.’” 

A final report with recommendations for lawmakers will be put together in coming weeks. 

Potential regulations and state uses 

The committee considered a number of proposals during the meeting for regulating the use of artificial intelligence, protecting consumers and exploring how the state government will interact with the existence of AI. 

One of the first proposed concepts was to apply existing state laws to AI models in the same way the laws apply to humans, including when it comes to disinformation, impersonation and creating and distributing fake imagery.

“Anything that’s illegal for a human, should be illegal for AI — as a general principle,” CEO of the MKE Tech Hub Coalition Kathy Henrich said during the meeting. 

Wisconsin has enacted two laws that regulate the use the AI already: 2023 Wisconsin Act 123, which requires disclosure of the use of generative AI in political ads, and 2023 Wisconsin Act 224, which created a new crime in the state prohibiting the possession of virtual child pornography and obscene material that could be created with generative AI.

Henrich acknowledged that this could be a complicated issue because any law would need to be able to determine a liable party. 

“Is it the developer? The deployer? The consumer?” Henrich said. “The devil will be in the details of making sure we’ve really thought through how you hold people accountable and where that accountability lies.” 

Sarah Alt, the chief process and AI officer for Michael Best, said the question of accountability for new technology is not a new dilemma, but it is complex because of how quickly AI has developed. 

Aly also proposed that the Legislature enact a couple of laws related to data privacy. 

Wisconsin considered a bill, 2023 Assembly Bill 466, that would have established certain requirements for “controllers and processors” of consumers’ personal data, including confirming what information they have, obtaining that data and requesting the deletion of data. It also included financial penalties of a forfeiture of up to $7,500 per violation. Three states, including Colorado, Connecticut and Utah, enacted similar data privacy laws in 2023.

Alt said that data is a “crucial raw material for artificial intelligence.” 

“If we’re not going to be able to legislate every possible use case… I do believe data is certainly one of the places where we would declare that to be no different than other raw materials that are regulated,” Alt said. She compared data to asbestos, saying, “there are safe ways that you can use asbestos and there are also very harmful ways that you can use asbestos.”

Jay Hill, Vice President of Advanced Technology for GE Healthcare, agreed that this type of regulation could help keep people accountable for how data is used.

“If it’s free, then people could do all sorts of crazy and exploitative things, but if there’s some cost to it, I think it elevates the level of responsibility for people who will use those data,” Hill said. 

Alt also said that lawmakers should look at resurrecting a bill similar to 2023 Wisconsin AB 824, which would have defined various roles and laid out a set of responsibilities for people or groups that own, control and share personal data. 

Members also discussed three other issues that lawmakers could explore including requiring the labeling of communications that use AI, possible bias embedded in the algorithms of large language models and requiring companies to publish clear AI guidelines and principles.

Anderson said that he thought that requiring disclaimers of AI when communicating with the public would be best practice. 

“I think I’m pretty good at discerning when AI is being used or not, but I think of my dad, who’s definitely not capable of determining what’s AI and what’s not… So, just making sure we’re protecting the public, so they’re aware,” Anderson said. 

Henrich cautioned that legislation such as that could potentially set a bad precedent by discouraging people from using their own discernment skills to determine whether AI was used. 

“If you legislate that you mark things as AI, people will get trained to believe that anything that’s marked as AI is AI, and anything that isn’t is human,” Henrich said. “If I’m a bad actor, I’m more likely to not market and therefore I’m more likely to fall for bad actors who may be using AI in the wrong ways.” 

Henrich said that lawmakers would also need to think about how that kind of legislation would be enforced.

Potential Investments in AI

The committee also considered potential areas where the state could invest to help support universities, technical colleges, businesses and other stakeholders in the ongoing education, innovation and integration related to AI.

One suggestion the committee could make is to expand Fast Forward, a program administered by the Wisconsin Department of Workforce Development that provides grants to employers to provide training to workers, and specifically dedicating funding to learning how to work with AI and requiring courses on computer science and statistics for students. 

Henrich said an investment in teaching people skills related to AI could be essential as the technology may affect the job market and could come through the Fast Forward program and through technical colleges and universities. 

“I’ve seen market dynamics in play, and as much as people want to say everyone will be employable long-term with AI, I am a firm believer that there’s disruption in the middle, that as AI is implemented normal market dynamics will go to saving costs and may result in people losing roles as a result of that,” Henrich said. “Investing up front and people understanding AI so that they maintain employable skill is going to be very critical.”

Henrich noted that there might need to be changes to the Fast Forward legislation to make it applicable to AI. 

Two potential suggestions the committee discussed involve incentives for businesses, including tax breaks and a grant or revolving loan program. 

Armstrong suggested that lawmakers could explore creating a program that would give small manufacturers an incentive to adopt automation and AI through a revolving loan fund or a small grant. He said he did a survey of small manufacturers in his part of the state and found that many aren’t really exploring the use of AI currently. 

“Only 4% of them were even looking at AI right now,” Armstrong said. “A lot of them, when I started talking to them [said] we don’t even know where to start at this point… They’re looking for somebody within a company to develop a curiosity about, what could AI do for us, and then I think you’d probably have a champion within that company that can run with it. This is more of a way to maybe jump start that.”

Alt said she liked the concept of a revolving loan because of how quickly technology moves and the fact that a company would have to keep working on improvements. 

“If your grant is one and done, then even if you do get what you needed that first time that you implemented it, there is still the constant tuning and the constant investing and the constant improving, and because the technology moves so quickly your idea is already old within months, if not a couple of years,” Alt said. 

The full list of potential recommendations to lawmakers can be found here.