Use of automated decision systems in employment settings regulated.
HF4445 would create a new set of worker protections governing the use of automated decision systems in Minnesota workplaces. The bill defines artificial intelligence, automated decision systems, electronic monitoring tools, worker data, and related terms, and then imposes notice, access, appeal, recordkeeping, and anti-retaliation requirements on employers that use these tools for employment-related decisions. It applies to a broad range of employment actions, including hiring, discipline, promotion, termination, scheduling, compensation, training, productivity requirements, and workplace health and safety.
Before using an automated decision system, employers would have to give advance written notice to affected workers, applicants, unions, and authorized representatives, and in many cases obtain affirmative written consent. The notice must explain the system’s purpose, data sources, logic, vendors, and worker rights. Employers must also keep records of worker data and system outputs, provide workers with access to those records, correct inaccurate data, and give post-use notice when an automated system is used in a decision. Workers would have a right to appeal decisions, and a human reviewer with authority to overturn the decision must review the appeal. The bill also prohibits certain uses of automated systems, including inferring sensitive traits, using facial/gait/emotion recognition, relying solely on automated outputs, and taking adverse action based on continuous time-tracking data except in limited cases.
The bill would add new sections 181.9921 to 181.9927 to Minnesota Statutes chapter 181, expanding state employment law to regulate employer use of AI and automated decision tools. It would impose new compliance duties on private employers, state and local government employers, labor contractors, and vendors involved in processing worker data, while creating enforceable worker rights to notice, access, correction, appeal, and protection from retaliation. The commissioner of labor and industry would be responsible for enforcement, and violations could lead to civil penalties, damages, injunctive relief, attorney fees, and joint and several liability for employers and vendors. The bill also preserves stronger local protections by stating that it does not preempt ordinances offering equal or greater worker protections.
The available record shows no committee transcript or vote history, so there is no direct evidence of debate, amendments, or recorded support/opposition. Based on the bill text alone, the measure appears strongly protective of workers and skeptical of unchecked automated decision-making, emphasizing transparency, human review, and limits on surveillance and algorithmic inference. The overall framing suggests a consumer- and labor-protection approach rather than a permissive regulatory model.
The main points of likely contention are the bill’s breadth and the compliance burden it places on employers and vendors. Employers may object to the advance notice and affirmative consent requirements, the detailed disclosure of system logic and impact assessments, the short deadlines for access and appeal responses, and the restrictions on using automated tools for compensation, discipline, and other employment decisions. Another likely area of dispute is the prohibition on certain data uses, including sensitive personal characteristics, facial/gait/emotion recognition, and continuous time-tracking data, as well as the bill’s joint and several liability provisions and civil penalties. Worker advocates and unions would likely support these provisions because they create meaningful rights to transparency, bargaining, and human oversight.