Minnesota AI Ethics Rules for Lawyers

The official instruments

A report asking a different question than everyone else

Most state instruments answer some version of “how may lawyers use AI ethically?” Minnesota’s asks who should get to use it at all. The MSBA AI Working Group Final Report, adopted by the Minnesota State Bar Association Assembly in summer 2024 and carried in our dataset at June 2024 with month precision, centers on the implications of large language models for unauthorized practice of law and access to justice. That is a different axis from competence-confidentiality-billing, and it makes Minnesota’s document the dataset’s clearest example of a bar thinking about AI as a distribution problem, who can deliver legal help and to whom, rather than purely a professional-risk problem.

Labeling discipline matters here. The document is a report, adopted by an assembly, and this tracker types it exactly that way; it is neither a formal opinion nor bar guidance dressed as one. Minnesota therefore lands in an in-between category that surveys tend to mangle: the state has a real, adopted instrument, but not one that reads out AI duties question by question. Both halves of that sentence are load-bearing, and the tracker preserves the distinction with type badges rather than rounding everything to “opinion.”

The sandbox recommendation and the standing committee

The report’s signature proposal is an AI regulatory sandbox: a framework under which nonprofits could deploy large language models to close the justice gap without UPL exposure. The idea takes the technology’s most promising public-interest use, cheap and scalable legal help for people who would otherwise get none, and pairs it with a supervised space in which UPL rules would not choke it off. A sandbox is a recommendation about regulatory architecture, not a permission slip; nothing in the report suggests nonprofits may start deploying today. But as a statement of direction from a state bar assembly, it is among the most forward-leaning positions in the dataset, and it treats access to justice as a reason to enable AI carefully rather than merely a risk to be managed.

Its companion recommendation is institutional: a standing MSBA AI committee, so the association’s engagement with the technology outlives one working group’s mandate. Recommendations of that shape are how a report becomes a pipeline, and it is the concrete thing to watch: committee output, sandbox movement, or a future opinion would each change Minnesota’s entry, and the changelog logs any verified development the week it lands.

What Minnesota lawyers actually practice under

For day-to-day duties, the report points back rather than forward: existing rules already require accurate outputs, confidentiality-preserving tools, and supervised delegation. Those three phrases compress the working compliance program. Accurate outputs means AI-assisted work gets verified before anyone relies on it, the same discipline that decides whether lawyers can safely use ChatGPT anywhere. Confidentiality-preserving tools means the platform’s handling of client information is examined before adoption, not assumed. Supervised delegation means AI work is overseen the way delegated human work is, with a responsible lawyer attached. Since Minnesota’s report does not itemize these duties further, ABA Formal Opinion 512 is the natural companion text for the detail, and the policy template turns the three-phrase summary into an adoptable firm document.

The report also leaves the courtroom untouched. Our dataset contains no statewide Minnesota court rule on AI in filings, which puts the filing-stage obligations where they sit in most of the country, in the general candor rules and in whatever standing orders individual judges maintain, federal courts included. Confirming the assigned judge’s orders before an AI-assisted filing is the remaining per-matter step, and the court orders guide shows what those orders typically require.