Somebody always asks what it costs. That is the right question, and it is where this kind of idea usually dies — not because the answer is bad, but because everyone in the room has watched an economic-development fund turn into a pile of money with a nice name on it.
So I will answer it first.
Nearly three years ago the federal government looked at Minnesota, found a real asset base, and said so in writing. Then it wrote the check to somebody else.
Minnesota has not done nothing, and pretending otherwise would sink this argument
The state is not standing still on artificial intelligence, and anyone who says it is has handed the other side an easy correction.
Minnesota IT Services runs a Technology Advisory Council, and in 2023 that council stood up an Artificial Intelligence Subcommittee — in the council's own words, to work "alongside the Transparent Artificial Intelligence Governance Alliance (TAIGA)." Its 2024 report describes what it actually does: build a strategic framework for AI adoption inside state agencies, support pilot projects, and work on bias, privacy, and governance recommendations.
And on January 3, 2025, the Governor announced a fraud package that put AI on the state's own books. His office's release is specific: "Use artificial intelligence to detect and prevent fraud. MNIT will partner with industry experts to create a pilot that would analyze data to help detect and flag payment anomalies for Medicaid providers through the Department of Human Services."
Both of those are real. Neither one is research funding. The first governs how the state buys and uses tools other people built. The second hunts thieves, which I am all for. Neither pays a single researcher in this state to work on the underlying problem — proving, on evidence, that an AI system does what its seller says it does.
Here is what survives that concession. No Minnesota session law in 2024, 2025, or 2026 appropriates a dollar for AI safety or verification research, and no council, task force, or center is charged with building that capacity. Nobody has walked into that field.
We already competed for this money, and we came in second
On October 23, 2023, the U.S. Economic Development Administration designated Minnesota MedTech Hub 3.0 one of 31 inaugural Tech Hubs under the CHIPS and Science Act. The EDA's own release says the hub will make Minnesota a global center for "Smart MedTech" "by integrating artificial intelligence, machine learning, and data science into medical technology." That is not an AI-adjacent designation. That is an AI designation, awarded by a federal agency running a competitive process, on the strength of assets we already had.
The same release told the winners what came next: eligibility for a phase that would "invest between $50-$75 million in each of 5-10 Designated Hubs."
Phase 2 landed on July 2, 2024. The EDA announced $504 million in implementation funding, in grants "ranging between $19 million and $51 million," to 12 of the 31 designated hubs, across projects serving fourteen states.
Minnesota is not on that list. Wisconsin is.
Twelve of thirty-one got funded and nineteen did not; reasonable panels make defensible choices. The point is that the recognition by itself bought nothing. A designation is a finding of fact. It is not a program. Something has to be built underneath it, and in three sessions since, nothing has been.
The number belongs in the bill. The structure belongs in the argument.
I am not going to tell you what this should cost. I do not know, nobody honest does yet, and a candidate who invents an appropriation is writing a press release rather than a budget. That figure comes out of a drafted bill, costed by nonpartisan fiscal staff, against a defined scope. Anyone who hands you a round number today is guessing at you.
What I will argue is what has to be true of the money before the number means anything.
Route it through research merit, not a legislative earmark. Money should go out through competitive, peer-reviewed award to the best proposal, not to a named center in a named district. A named earmark is how economic development becomes patronage, and how these funds rot.
A quirk in Minnesota law helps here. Our constitution says, in a single sentence, that "all the rights, immunities, franchises and endowments heretofore granted or conferred upon the University of Minnesota are perpetuated unto the university." That has long been understood to mean the Legislature funds and contracts with the University rather than commands it. Which cuts one way for a bill drafter: the conditions ride on the appropriation or they do not exist. Write them in.
Publish everything. What was proposed, who was funded, how much, and what came out — including the methods that failed. A verification technique nobody outside the lab can read is not a verification technique, and a grant nobody can audit is a grant nobody will trust. Publication is not a courtesy here. It is the product.
Measure it in results, not ribbon-cuttings. Count published methods, credentials issued, students placed, firms that opened here. Not announcements, not groundbreakings, not the number of times a press release used the word "innovation."
Sunset it. Put an expiration date in the statute, and make renewal require a showing against that published record. A program that cannot be killed will outlive its usefulness by a decade, and every legislator reading this knows one.
If the benefit only lands downtown, the rest of the state paid for a press conference
Most of the work in this field is not doctoral work. Somebody has to run the test suite, maintain the evaluation corpus, document where a model file came from, check a signature, and write the finding up in a form that holds up when a lawyer cross-examines it. That is a two-year credential and a career.
Minnesota State describes itself as "the third largest system of state colleges and universities in the United States and the largest in the state with 26 colleges, 7 universities, and 54 campuses," serving "more than 290,000 students each year." That is the system's own description, which is why I quote it rather than characterize it.
Fifty-four campuses is a delivery network. If the training and certification run through it, this field means a job in Hibbing and a job in Marshall. If it runs only through one downtown, then greater Minnesota helped pay for something it will read about.
Ethics written only by engineers and lobbyists is ethics nobody trusts
Whatever body writes Minnesota's AI standard should have citizens sitting on it. Not a comment period. Seats.
Public input is slower and messier than a working group of experts. But a standard about whether a machine can be trusted to decide something about your loan, your benefits, your kid's school placement, or your medical care is not a technical document. It is a public one. Written only by the people who build and sell these systems, it is a marketing brochure with a state seal on it, and the buyers we want — hospitals, banks, school districts, the state itself — will price it accordingly.
The concession, and I am not going to wave it away
States have wasted staggering sums chasing industries. Texas spent to win SEMATECH in 1988 and watched it move its headquarters to Albany in 2007 after New York outbid it. That is the pattern, not the exception, and a Minnesota AI fund is exactly the kind of program that becomes a slush fund — a friendly appropriation, a favored institution, a decade of annual renewals, and no one able to say what it produced.
I cannot tell you that merit review, publication, and a sunset eliminate that risk. They do not. They manage it. They make the failure visible early and cheap to stop, which is the most any structure can do to money. Anyone promising you a guarantee on this is selling you the program.
Every mechanism I have listed makes this fund easier to kill. That is the feature, not the cost.
What we can do
Appropriate it competitively, never by earmark. Peer-reviewed award to the best proposal; no named institute, no named district.
Write the transparency into the appropriation. Proposals, awards, methods, results, and failures published as a condition of the money.
Sunset it, and make renewal earn itself against the published record, not against a lobbyist's memory.
Run the training and certification through the Minnesota State campuses, so the credential reaches all fifty-four of them and the benefit is not one zip code's.
Put citizens in the seats on the standard-writing body, not in the comment file.
I laid out where I stand here. This is the plumbing underneath it, which is where public money is either respected or wasted.
A fund with no sunset and no published record is not an investment. It is a subscription.
First the facts. Then the fix.
Sources
The MNIT facts come from the Technology Advisory Council's own 2024 annual report, pulled as a raw PDF from mn.gov and read directly: the council "prioritized AI as a critical area of growth in 2023," which "culminated in the establishment of the TAC's AI Subcommittee to work alongside the Transparent Artificial Intelligence Governance Alliance (TAIGA)," and the subcommittee's described work is agency adoption, pilots, bias, privacy, and governance recommendations. The fraud-detection language is quoted verbatim from the Governor's own press release, "Governor Walz Takes Executive Action, Announces Legislative Package to Stop Fraud," dated January 3, 2025, retrieved through an Internet Archive snapshot of the mn.gov page because the live page does not serve its text to automated retrieval. The Tech Hubs facts come from the EDA's own documents, both retrieved through Internet Archive snapshots of the identical eda.gov URLs (eda.gov serves a bot challenge): the October 23, 2023 designation release for "Minnesota MedTech Hub 3.0," the "Smart MedTech" and "artificial intelligence, machine learning, and data science" language, and the "$50-$75 million in each of 5-10 Designated Hubs" statement of what Phase 2 would be; and the Tech Hubs Phase 2 Portfolio Fact Sheet for the July 2, 2024 award of $504 million in grants "ranging between $19 million and $51 million" to 12 of the 31 designated hubs across fourteen states, a list that does not include Minnesota. The constitutional sentence is quoted verbatim from the Minnesota Constitution, article XIII, section 3, pulled raw from the Revisor of Statutes. The Minnesota State figures are quoted from Minnesota State's own system page and are that system's description of itself, retrieved directly. The finding that no Minnesota session law in 2024, 2025, or 2026 appropriates money for AI safety or verification research comes from a source-verification pass across those three sessions' laws; the Revisor's chapter lists for all three years contain no AI chapter, and the one on-point bill of the 2026 session — H.F. 4544 and its Senate companion S.F. 4636, licensing private AI verification organizations — was introduced March 23, 2026 and died in committee without enactment. SEMATECH's 1988 arrival in Austin and its 2007 headquarters move to Albany after New York outbid Texas are drawn from the research record on Austin's technology consortia.
The negative finding — that no appropriation for AI research exists — was verified against session-law chapter lists and a research pass, not by reading every omnibus appropriations article line by line; if a line item is buried in one, I want to know about it. The MNIT Medicaid fraud pilot is quoted as announced in January 2025; whether it is operational today was not verified here. The proposition that the Legislature may fund and contract with the University of Minnesota but not direct it rests on judicial interpretation of article XIII, section 3 that was not independently verified for this piece — the constitutional text itself is quoted exactly. Minnesota State's "third largest system" claim is its own and is not independently checked. Corrections: campaign@madgettformn.com.