In early 2025, I encountered a high school English teacher at the dog park, someone I have known for many years, who was lamenting the educational crisis of the moment: artificial intelligence and the epidemic of students passing off AI-generated work as their own.

“These kids don't have any interest in learning anything,” she fumed. “Even ones who should know better. They all tell me, 'Why bother? In a few years, AI will just do it all anyway.'” She seemed not just aggravated, but despondent.

I knew this person to be an engaged, caring, thoughtful educator. She has an elite academic pedigree, professional IT expertise, and a wealth of hard-earned experience in the classroom. On the subject of AI, however, she was sounding as jaded and burned out as Springfield Elementary's Edna Krabappel (may she, and the actor who voiced her, rest in peace).

Thinking I could help, I offered observations from the work I had been doing with generative AI in the legal profession: the hazards of relying on large language models without checking their outputs, the importance of clear operational constraints, the value of retrieval-augmented systems, and so on.

None of it made a dent. She remained focused on her understandable worries: that a whole cohort of students was passing through high school learning mainly how to game the system; and that the time she spent chasing down AI-generated content was time she couldn't spend teaching, giving meaningful feedback, or just having a normal weeknight (a reasonable request for both teachers and students alike).

After a few minutes, it was clear that we were talking past each other and my “helpful” suggestions were only adding to her irritation, so I let it drop.

I was reminded of this conversation recently when looking at a State of Michigan job posting sent to me by a colleague. I navigated to the state's talent acquisition portal hosted by NEOGOV, a vendor whose homepage advertises its product as “AI-Enabled Government HR Software” and invites prospective clients to “Streamline everyday government HR processes with an AI-supported platform.” There, plastered prominently across the job posting was the following:

The use of artificial intelligence (AI) software of any kind is prohibited in all areas of the selection process; including, but not limited to, responses to application questions, and responses to interview questions or exercises. Any applicant suspected of using AI software for any portion of the selection process may be screened out.

(Emphasis in original.)

One wonders if the irony of this boldfaced, one-sided warning crossed the minds of whoever made the policy or approved the posting—to say nothing of its probable inaccuracy. It is hard to believe that AI is “prohibited in all areas of the selection process” by the department in question; in practice, it is more likely to be prohibited only for applicants. The human resources office, I would venture, is free to use at least some of the AI-supported tools available under the state's contract with NEOGOV.

Aside from being taken aback by the rather glaring hypocrisy, I had to wonder: why was this department—one which has a major role to play in Michigan's legal and economic systems—so eager to play the unwinnable game of AI whack-a-mole that was burning out my teacher friend?

Narratives that treat AI's future omnipresence as a fait accompli are justifiably becoming unpopular outside of the tech world. I am sympathetic to those who have called for industry to stop treating every new AI advance as an inevitability, and start thinking about how it benefits society beyond “enhancing shareholder value.” To ask, in other words, not only whether they could, but if they should.

Nonetheless, it is hard to ignore the strategic value proposition of AI when it performs as advertised. Moreover, the likelihood that multinational corporations (or foreign adversaries, or students, or job applicants) will unilaterally disarm themselves is vanishingly small.

Prohibition is therefore not a strategy. As a policy, it functions as a temporary avoidance mechanism, creates a futile administrative burden, and makes a poor substitute for clear, enforceable norms. A better approach for all of us, including government agencies, is to figure out how to use and enhance AI as a tool for good, rather than as an end in itself.

The author has been a licensed Michigan attorney since 2002, with professional experience in multiple public sector roles.

This is Part One of a three-part series on how Michigan's public sector is responding to AI. Part Two will look at the legal profession—closely tied to the public sector, already forced to wrestle with AI's impacts, and perhaps pointing the way to better policy approaches.

Full disclosure: AI didn't write this piece, but it did help refine it.