Artificial intelligence creates a brave new world for state agencies in Michigan

This story was originally published by Michigan Advance.

As artificial intelligence increasingly seeps into public life, the question of if – and how – governments should incorporate this technology into their daily operations looms large over Michigan.

While state lawmakers have set regulations barring political campaigns from using promotional materials generated with artificial intelligence without a disclosure, legislation looking at how Michigan’s 19 state agencies use artificial intelligence is just beginning to see movement.

At the end of April, state Rep. Jaime Greene (R-Richmond) introduced a bill to create an artificial intelligence governing board within the Michigan Department of Technology, Management and Budget, the agency tasked with overseeing technology use for all other departments. 

The current version of the bill states the board would consist of three members appointed by the governor: one individual with expertise in artificial intelligence or data science, one individual with expertise in ethics, civil rights, or privacy, and one individual who represents the private sector.

Alongside advising the department on proposed uses for generative artificial intelligence and providing periodic updates to the department’s formal guidelines for agencies incorporating AI into their operations, the board would also assist in creating a pilot program to study the use of generative AI in state government.

Greene told Michigan Advance she’d already begun working on her House Bill 5899 when the Department of Technology, Management and Budget released its guidelines on AI use in 2025, and that the document aligned closely with her vision for the legislation.

Although AI technology is not new, the department’s guidance acknowledges these tools have evolved significantly, presenting opportunities for the state to improve existing digital services for Michigan residents and offer new products, solutions, services, and programs. It also notes that these new capabilities come with significant responsibilities for the agencies and any employee who makes use of them.

Speaking to members of the House Communications and Technology Committee, Greene noted that there is nothing preventing any state agency from using AI in their work now, and that the pilot program would allow state officials to see how departments are using this technology. 

“This is a great opportunity to experiment with AI in a controlled setting,” Greene said. The committee sent the legislation to the House Rules Committee on Tuesday.

Within 180 days of the program’s completion, the Department of Technology, Management and Budget would be required to submit a public report evaluating any time saved and processes made more efficient alongside any risks or unintended consequences they identified from generative AI use.

The bill also states that any generative AI system within the program must not be used to carry out or enable an activity that violates the law, discriminate in violation of the law, adversely affect an individual’s privacy rights, or allow the unauthorized access or use of personal data or personally identifiable information.

Program participants would also be required to ensure that any output from the generative AI program under study is reviewed by a human.

“Human review is really, right now, our strongest accountability in AI,” Greene said.

While working on her legislation, Greene said the departments did not reveal how they were using AI in their work, but cited the existence of the department’s guidelines as proof that they are using it.

What Does Responsible AI Use Look Like for State Governments?

As Greene’s bill awaits further action within the House Rules committee, the question of how state agencies are using artificial intelligence remains, particularly given the state’s history with AI. 

In 2013, the state sought to automate the review of unemployment cases through the Michigan Integrated Data Automated System, which falsely accused more than 40,000 individuals of fraud within its first two years of use. The system sparked multiple lawsuits over financial harms and the lack of human review of its determinations.

In a review of 22,000 cases marked as fraudulent, the Michigan Auditor General determined that 93% did not actually involve fraud.

Molly Kleinman, the managing director of the Science, Technology, and Public Policy program at the University of Michigan’s Ford School of Public Policy said while state governments are trying to be deliberate about their approach to AI adoption, there is a lot of experimentation on the individual level. 

“It’s not across the board by any stretch, but pretty much in the absence of a clear policy saying ‘don’t do it,’ or ‘here’s how you can do it,’ people are trying stuff,” Kleinman said. 

One of the big challenges that comes with governments regulating technology is the fact that it is accessible to employees at the individual level, Kleinman said, noting that anyone can make a free ChatGPT account, or pay for a Claude subscription.

Data privacy and data security are chief among the concerns that governments should be looking at when considering their AI use, Kleinman said. 

Additionally, AI is a marketing term that applies to a wider range of software, including programs we would have referred to as machine learning or algorithms in the past, Kleinman said. As such, issues like algorithmic bias – where systems produce repeated, harmful outcomes – are still present, Kleinman said.

Additionally, the environmental footprint of technology is another piece worth considering, Kleinman said, noting the bipartisan pushback against data centers used in AI development. 

In addressing concerns around data security, government officials need to negotiate questions of who has access to search terms, results, and anything else they enter into the system, Kleinman said, noting that many are approaching this issue by looking at how a state acquires technology. 

“How do we need to change the way a state approaches procurement in order to successfully mitigate the threats of these AI systems and also reap the benefits that might be there?” Kleinman asked. 

If states want to stay on top of any possible benefits, flexibility will be key, Kleinman said, noting that as AI companies form, merge and dissolve, there needs to be some ability to pick programs up and let them go. 

Flexibility is also important on the management side for ensuring employees are comfortable experimenting with AI, and that they’re being transparent about their use, Kleinman said.

“People are trying these systems, it’s worse if they aren’t telling you,” Kleinman said. “So to encourage people to try stuff out, be open about what works and what doesn’t, be tracking when there are problems and errors and mistakes. Have a very clear system for tracking those things, one that ideally is not going to be punitive on the staff.”

When looking at specific applications for AI within government, Kleinman said public-facing systems are at the end of her list for where these systems should be implemented, and are often the first place governments look. 

Michigan’s AI strategy notes that customer service chatbots, which are designed to simulate conversation with humans, were highly requested by state agencies.

Interacting with these bots often triggers frustration among residents and can damage public trust in the long-term, Kleinman said, pointing to systems built for a specific purpose and trained on specialized sets of data as a significantly more promising use for governments.

“I’ve heard some great excitement around transit optimization systems,” Kleinman said, “Something that these tools are generally good at is pattern identification; finding patterns that it would be much harder for a human to identify, because it’s just the scale of the math that these systems can do.”

This type of program can help examine questions like how to improve a transit system, whether or not another bus can fit onto a route, or whether another route can fit into a specific part of a city, Kleinman explained.

Questions of Fairness and Accountability Persist

Taking a broad look at the regulatory landscape for AI systems across the country, Kleinman warned that states are woefully behind on enacting measures to ensure these programs are used safely. 

“We’re behind, and we’re behind in ways that in some ways we’re not even talking about, because really we should be regulating these companies,” Kleinman said. “The kinds of harmful tools that are reaching the public should not be. And if we were regulating the companies at all, they wouldn’t be. And then we wouldn’t be leaving it to every municipality, every state, every department to figure these things out for themselves.”

While conversations around the negative aspects of AI often focus on “big future apocalypses” these discussions distract from the real harms that are already present, Kleinman said. 

“The patterns that we see over and over and over again across many technologies, across many decades is that the people most likely to be harmed by a new technology are the people who are already marginalized,” Kleinman said. “Whether that’s poor people, people of color, immigrants, children, people with disabilities, it’s over and over and over again we see the same patterns. And these AI systems, in particular, because of how they were trained, are sort of uniquely positioned to exacerbate a lot of the inequalities that already exist in our society.”

In 2024, the Michigan Civil Rights Commission passed a resolution adopting its own guiding principles on AI use, which includes provisions stating Michigan residents should not face discrimination by algorithms and systems, should be protected from abusive data practices and should be aware when an AI-powered system is in use.

The Michigan Advisory Committee to the United States Commission on Civil Rights is also moving to study the civil rights implications of AI in Michigan.

When defining AI within government, Kleinman pointed to the Blueprint for an AI Bill of Rights introduced by the White House Office of Science and Technology Policy in 2022, which describes AI as automated systems that have the potential to meaningfully impact the American public’s rights, opportunities, or access to critical resources or services.

This definition does not focus on the specifics of a program or system, but rather the social, ethical and political impacts that come with this technology, Kleinman said. 

“Whenever you’re talking about government use of AI, I think it’s really important to start with ‘how might this impact people’s rights?” Kleinman said.

Michigan Advance is part of States Newsroom, a nonprofit news network supported by grants and a coalition of donors as a 501c(3) public charity. Michigan Advance maintains editorial independence. Contact Editor Jon King for questions: info@michiganadvance.com.