License for artificial intelligence independent verification organizations established, advisory council established, rulemaking authorized, and reports required.
HF4544 creates a new regulatory framework in Minnesota for “independent verification organizations” (IVOs) that would be licensed by the commissioner of commerce to assess artificial intelligence models and applications. The bill defines key AI-related terms, requires an IVO to submit a detailed proposed plan showing how it will evaluate and mitigate specified risks, and authorizes the commissioner to license the organization only if the plan is adequate and the organization is sufficiently independent from the AI industry. The bill also allows the commissioner to limit a license to particular risks or market segments and to revoke a license if the IVO’s plan becomes misleading, outdated, compromised, or materially ineffective.
Once licensed, an IVO must implement its approved plan, monitor verified AI systems over time, revoke verification when developers or deployers fail to meet mitigation or governance requirements, and report material plan changes to the commissioner. The bill also establishes an Artificial Intelligence Advisory Council within the Department of Commerce, with at least one civil society representative, and gives the commissioner authority to delegate licensing and auditing functions to the council. The council members are subject to conflict-of-interest restrictions, term limits, and postemployment limits intended to preserve independence.
The bill further requires annual reporting to the commissioner and legislative committees, with redacted reports posted publicly by the Department of Commerce. Those reports must include aggregated information on the capabilities, risks, benefits, verification results, remediation compliance, and any additional significant risks observed during assessments, while protecting trade secrets, sensitive business information, personally identifiable information, and security-sensitive content. The bill also requires retention of verification-related records for ten years.
A notable legal effect of HF4544 is its creation of a rebuttable presumption against liability in civil actions for personal injury or property damage when an AI system was verified by a licensed IVO for the relevant risk and market segment at the time of the alleged injury. In practice, that provision would give verified AI developers and deployers a potential defense in lawsuits, while still allowing plaintiffs to overcome the presumption with contrary evidence. The bill also authorizes the commissioner to adopt rules to implement the new chapter provisions and states that verification is voluntary, not mandatory.
HF4544 would add a new chapter to Minnesota Statutes governing AI verification and oversight, centered in the Department of Commerce. It would create licensing, reporting, recordkeeping, and enforcement authority over private independent verification organizations, while indirectly affecting AI developers and deployers by imposing mitigation, monitoring, governance, and corrective-action expectations on systems that seek verification. The bill also introduces a civil-liability presumption that could affect tort litigation involving AI-related personal injury or property damage claims, especially for systems that have been verified for the relevant risk.
No committee transcripts or recorded votes were provided, so there is no direct evidence of legislative debate or formal support/opposition in the materials supplied. Based on the bill text alone, the measure appears designed to balance innovation with consumer protection by creating a structured, state-supervised verification regime for AI systems. The overall tone is regulatory and precautionary, with an emphasis on independence, transparency, and ongoing monitoring.
The main likely points of contention are the scope and practicality of state oversight of AI, the cost and burden of licensing and ongoing reporting, and whether the proposed liability presumption gives verified companies too much protection in court. Another possible issue is the breadth of the commissioner’s and advisory council’s authority to define acceptable risk, approve or reject verification plans, and require corrective actions. Stakeholders most likely to support the bill are consumer advocates, civil society groups, and policymakers concerned about AI safety, while AI developers, deployers, and industry groups may be concerned about compliance costs, confidentiality, and the potential for regulatory uncertainty.