Minnesota Bar Briefing September 2026 Edition

The Minnesota Dual-Track Model:
The Regulatory Sandbox vs. The Sanctions Hammer

How Minnesota became the primary laboratory for legal AI regulation: championing algorithmic safe harbors for civil legal aid while enforcing strict non-delegable supervisory liability under MRPC Rule 5.3 and federal Rule 11.

Why We Compiled This: Our team at Koce builds computational legal analysis and trial-prep workflows for litigators across the country. In following the recent decisions coming out of the Minnesota Supreme Court and the MSBA AI Standing Committee, we realized there wasn’t a single, transparent briefing that clearly synthesized both sides of the coin: the bar’s pro-access innovation efforts and the OLPR’s strict disciplinary enforcement. We compiled this neutral briefing for Minnesota attorneys to provide a clear, practical roadmap. We hope it’s helpful.

The Core Dynamic

Two Competing Realities Shaping Minnesota Legal Practice

Track 1: Pro-Innovation

The MSBA AI Regulatory Sandbox

The Minnesota State Bar Association’s initiative to bridge the civil justice gap by carving out an algorithmic safe-harbor under Minn. Stat. § 481.02, expanding the Supreme Court’s Supervised Practice framework.

Access to Justice Target: Deploys automated tools in high-volume, pro se dockets (Hennepin Housing Court eviction defense, immigration asylum, debt claims).
The Paraprofessional Precedent: Modeled after Supreme Court Order No. ADM19-8002, legally treating AI as a supervised "authorized agent."
Attorney-First Guardrails: Rejects Utah/Arizona pure corporate ownership; preserves the licensed Minnesota attorney as the sole nexus of legal accountability.

Track 2: Disciplinary Enforcement

The OLPR Supervisory Standard

The Minnesota Office of Lawyers Professional Responsibility and state/federal courts applying strict liability to AI citations under MRPC Rule 5.3 (Nonlawyer Supervision) and Federal Rule 11.

The Nonlawyer Supervision Doctrine: Autonomous LLMs are legally classified under MRPC 5.3. Software failure is treated as an attorney supervisory breach.
Kohls v. Ellison Precedent: The District of Minnesota struck an entire expert declaration for citing hallucinated sources, extending Rule 11 verification to expert oversight.
No "Algorithm Defense": Blaming an AI tool for a fictitious citation is considered an independent breach of Candor to the Tribunal (MRPC Rule 3.3).

92% Civil Justice Gap Low-income Americans receive no or inadequate legal help for 92% of civil legal problems (LSC).

MRPC 5.3 Governing Ethics Rule Applied by OLPR Director directives to non-human algorithmic drafting and research.

Kohls v. Ellison Foundational Case Federal Rule 11 standard barring submission of unverified AI expert declarations.

Zero Retention Confidentiality Rule MRPC 1.6 requires verified enterprise agreements that prohibit training on client inputs.

Track 1 Analysis

The MSBA Sandbox Architecture: Solving the § 481.02 UPL Friction

The structural foundation of Minnesota’s pro-innovation track was formulated by the MSBA Artificial Intelligence Working Group in its landmark June 2024 report, Implications of Large Language Models on the Unauthorized Practice of Law and Access to Justice, and advanced through the permanent MSBA AI Standing Committee established under the leadership of Damien Riehl. The Working Group recognized a critical shift in modern legal practice: the traditional boundary between permissible "legal information" and regulated "legal advice" has become largely indistinguishable when using advanced Large Language Models. Because modern models synthesize facts and apply statutory law directly to complex fact patterns, consumer-facing legal technology inherently implicates unauthorized practice of law (UPL) statutes.

Under Minnesota Statute § 481.02, Subd. 1, it is unlawful for anyone other than a licensed attorney to practice law or give legal advice. To deploy AI in underserved dockets—such as Hennepin County Housing Court or conciliation claims—without facing criminal or civil UPL enforcement, the MSBA proposed an innovative structural solution: rather than seeking blanket deregulation or legislative revisions to criminal UPL statutes, establishing an AI Regulatory Sandbox that analogizes AI tools to human nonlawyers operating under the Supreme Court's established Supervised Practice framework.

"Rather than seeking blanket deregulation or legislative revisions to criminal UPL statutes, Minnesota’s emerging safe-harbor approach bridges the justice gap by analogizing AI models to human nonlawyer assistants operating under the direct supervision and responsibility of licensed counsel." — Regulatory Analysis | Koce Intelligence

By modeling the proposed AI Sandbox on the Minnesota Supreme Court’s Legal Paraprofessional Pilot Project (Order No. ADM19-8002)—which the Court successfully made permanent effective January 1, 2025—the framework establishes an "authorized agent" doctrine. Under this model, an AI tool functions as an instrument of a licensed supervising attorney, ensuring that:

Client confidentiality and privileged work product remain protected under MRPC Rule 1.6 and attorney oversight;
The supervising lawyer assumes personal, nondelegable professional liability under MRPC Rule 5.3 for any algorithmic output;
Direct courtroom advocacy, binding settlement execution, and final strategic counseling remain reserved exclusively for licensed counsel.

Sandbox Status Update: On August 27, 2025, the MSBA AI Standing Committee formally approved and submitted its recommendation to the Minnesota Supreme Court requesting an administrative order establishing the AI Regulatory Sandbox. As of 2026, the proposal remains under consideration by the Court.

Track 2 Analysis

The Disciplinary Anatomy of "Hallucinations": Kohls v. Ellison & Rule 5.3

While the MSBA pursues structural access, the Office of Lawyers Professional Responsibility (OLPR) has established clear disciplinary guardrails. In formal policy directives, the OLPR clarified that Minnesota does not require brand-new disciplinary rules to govern AI; existing rules—specifically MRPC Rules 1.1 (Competence), 1.6 (Confidentiality), 3.3 (Candor to the Tribunal), and 5.3 (Supervision of Nonlawyers)—are fully enforceable against AI misuse.

This standard was tested in federal court in Kohls v. Ellison (U.S. District Court for the District of Minnesota). In that case, an expert declaration submitted on behalf of the Minnesota Attorney General’s office was found to contain fabricated citations generated by an LLM. The Court took immediate action:

Striking the Evidence: The entire expert declaration was stricken from the record, critically undermining the evidentiary basis of the filing.
Personal, Nondelegable Oversight: The Court ruled that attorneys hold an individual, nondelegable responsibility under Rule 11 to independently verify every factual and legal proposition, extending this duty explicitly to questioning their own retained expert witnesses about whether generative AI was used.
The Ethics Implication: The OLPR has emphasized that attempting to deflect blame to an algorithm is treated as an aggravating factor under MRPC Rule 3.3 (Candor) and Rule 8.4(d) (Conduct Prejudicial to Justice).

National Benchmarking

Comparative Analysis: Minnesota vs. Texas, Florida, and the ABA

To understand where Minnesota sits in the broader national picture, it helps to compare its dual-track standard against the country's most prominent regulatory frameworks:

Jurisdiction	Verification Standard	Confidentiality & Privacy	Fee & Billing Regulations	Client Disclosure
Minnesota	Nondelegable verification (Kohls); strict supervisory liability under MRPC 5.3.	Informed consent required for non-ZDR cloud tools; boilerplate consent banned.	Governed by MRPC 1.5 reasonableness; efficiency gains must benefit client.	Affirmative duty to disclose under MRPC 1.4 if AI use is material to matter.
Texas	Independent human verification mandated (Ethics Op 705).	Strict confidentiality via enterprise tool vetting.	Strictly bars billing for AI time saved; only actual time worked.	No mandatory disclosure unless costs are passed through.
Florida	Independent oversight mandated (Advisory Op 24-1).	Must ensure confidential data is shielded from public training.	Reasonable fee requirement; prohibits duplicate charging.	Standard communication duties under Rule 4-1.4 apply.
ABA Model	Must maintain "reasonable understanding" (Formal Op 512).	Must protect client data from unvetted public platforms.	Cannot bill client for time spent learning the tool.	Disclosure required if asked or if material to representation.

Practical Compliance

Law Firm AI Compliance: A 4-Step Operational Checklist

For Minnesota managing partners, litigators, and corporate counsel, operating safely within this dual-track environment requires concrete internal protocols:

Step 1: Acceptable Use Policy

Document Firm AUP

Prohibit the input of client facts, deposition transcripts, or medical records into consumer, public-tier LLMs (e.g., free web chatbots) where prompts are stored for training.

Step 2: Vendor Due Diligence

Require ZDR Guarantees

Ensure all commercial software vendors contractually provide Zero-Data-Retention (ZDR), SOC 2 Type 2 verification, and explicit indemnification against data training.

Step 3: Nondelegable Verification

Enforce Manual Checks

Every citation, statutory deadline, and case quotation must be verified against primary reporters before filing. Never rely on an LLM's own self-verification.

Step 4: Client Communication

Specific Informed Consent

Per OLPR guidance, avoid vague engagement clauses. If non-ZDR tools are utilized, provide clear disclosures regarding third-party data transmission.

Access to Justice Initiative

Koce Access to Justice (A2J) Commitment: 100 Pro Bono Matters

Technology should close the civil justice gap, not widen it. To directly support Minnesota legal aid organizations, pro bono litigators, and the emerging regulatory sandbox ecosystem, Koce has established an initial 100-Matter Access to Justice Allocation:

100% Subsidized Computational Workflows: Fully free access to Koce’s record-grounded litigation analysis, document extraction, and timeline synthesis workflows for up to 100 qualifying civil matters.
Partner Eligibility: Available to attorneys and staff at recognized Minnesota 501(c)(3) civil legal aid organizations (including Mid-Minnesota Legal Aid, Southern Minnesota Regional Legal Services, and Volunteer Lawyers Network) and court-approved Legal Paraprofessionals operating under attorney supervision.
Supervised Practice Integrity: Because Koce is a technology provider and does not directly practice law, every participating matter must be directed by a licensed Minnesota supervising attorney under MRPC Rule 5.3.
Zero Data Retention Guarantee: Full compliance with MRPC Rule 1.6 confidentiality. Client materials and uploaded exhibits are never used to train machine learning models.