Newly proposed legislation in Michigan would impose burdensome requirements on artificial intelligence systems. The predictable impact from these proposed laws would be to drive AI development out of the state, likely to China or other Asian economies.

House Bill No. 4667 would create new felony offenses and mandatory prison sentences for the criminal use, development, or distribution of AI systems. House Bill No. 4668 would require AI developers in Michigan to conduct regular risk assessments and third-party audits, as well as implement and publicly disclose safety protocols. Both bills were introduced by Rep. Sarah Lightner, R-Springport. HB 4668 has been referred to the House Communications and Technology Committee. HB 4667 has been sent to the House Judiciary Committee.

A notable feature of HB 4667 is that it would establish an absurdly broad definition of AI as “any machine-based system that can process data, generate content, or simulate human-like interactions, including, but not limited to, chatbots, voice assistants, generative AI models, and automated decision-making tools.” By this definition, basic spreadsheet sort functions would qualify as an AI system. So would routine business marketing analytics, and almost any customer profiling using a computer. Many small business owners have no idea that when they maintain and sort routine customer data in their Excel spreadsheet, they are using “AI.” Yet if they are prosecuted for a separate offense, even a minor violation, they could face a mandatory eight-year prison sentence when the prosecutor adds a violation of HB 4667 to the charges.

An overly broad definition of AI combined with heavy-handed regulations “could inadvertently impose stringent regulatory obligations on common practices that have minimal, if any, adverse impacts on consumer welfare or privacy interests,” according to a recent article by the International Center for Law and Economics. “Small and medium-sized enterprises, in particular, would face significant uncertainty and disproportionately high compliance burdens,” it said. “Indeed, smaller firms already rely heavily on low-risk AI applications to boost productivity and maintain competitiveness in an increasingly technology-driven marketplace.”

The Michigan Chamber of Commerce has concerns, too. “While Michigan HB 4668 is well-intentioned in seeking to address real concerns around AI misuse, it places overly burdensome and impractical requirements on developers. The MI Chamber emphasized that rather than having individual states pass their own AI regulations — risking a patchwork of conflicting and duplicative rules across the country — this issue should be addressed through federal legislation.”

These bills create penalties for things that are already illegal. Frauds, scams, defamation, illegal discrimination, and anti-competitive business conduct violate the law regardless of whether AI was used in the process. Thus, it is not clear how victims of illegal conduct will benefit from these proposed AI laws. One group that would benefit greatly, though, is trial attorney bar. Some attorneys would reap new business opportunities by defending clients against prosecutions under HB 4667. Others would gain by bringing class action lawsuits for failing to comply with all the new regulatory requirements under HB 4668, which would apply regardless of whether anyone is actually harmed.

HB 4667 and 4668 follow the regulatory approach of President Joe Biden’s much-criticized 2023 AI executive order, which in the words of analyst Kristin Stout, “sees dangers around every virtual corner” and imposes “regulations born of fear [that] threaten to derail beneficial innovation.” The Biden AI executive order states that the administration “places the highest urgency on governing the development and use of AI safely and responsibly.” President Donald Trump rejected this kind of government-directed approach to AI development when he repealed the Biden AI executive order in his first week in office.

On July 23, Trump issued his own AI executive order, “Winning the Race: America’s AI Action Plan.” The new executive order stresses that cybersecurity, technological innovation, and infrastructure construction are not opposing or wholly separate efforts, but rather complementary objectives to be achieved in tandem. The proposed Michigan laws, by contrast, fail to take into account the importance of promoting technological innovation. Instead, they focus solely on imposing regulatory restrictions and punitive punishments.

American companies are already struggling with excessive state and local mandates, dubbed the “Sacramento Effect” for California’s misadventures in regulating AI. This places our most innovative companies at a competitive disadvantage with foreign competitors and undercuts efforts to stay ahead of China in the race for AI supremacy. HB 4667 and 4668 would add an unwelcome “Lansing Effect” to the most innovative companies in Michigan and give them an incentive to relocate to a more welcoming state, or even another country.

Permission to reprint this blog post in whole or in part is hereby granted, provided that the author (or authors) and the Mackinac Center for Public Policy are properly cited.