MN SF4380

Online platforms conducting business in Minnesota that use algorithmic recommender systems based on personal data must publicly disclose a list of all algorithms used, descriptions of inputs and data sources, and weights categorized by quartile importance

Algorithmic recommender systems must be configured by default to maximize long-term user value metrics rather than engagement, and platforms must provide accessible interfaces for users to set preferences on recommended and blocked content types

Platforms must maintain holdout groups for at least 12 months to assess algorithm design changes and publish annual disclosures of long-term user value measurements, with mandatory independent audits at the business's expense

Special protections apply to minors, requiring algorithms to default to maximizing long-term user value metrics applicable to minors when a platform knows or should know a user is under 18

Violations constitute unfair and deceptive acts enforceable by the Attorney General, with civil penalties of $5,000 per user per violation (or $7,000 for reckless/knowing violations), plus attorney fees and injunctive relief