This bill creates new rules to protect people from harmful artificial intelligence systems used in important life decisions. It applies to AI algorithms that affect things like job hiring, school admissions, housing rentals, loans, insurance, voting, criminal justice, and government benefits.

Before companies can use these AI systems, they must evaluate whether the algorithm could cause harm or unfairly discriminate against people based on protected characteristics like race, gender, disability, or national origin. If there's a real risk of harm, an independent auditor must conduct a detailed review of how the AI works, what data it uses, and whether it treats different groups fairly. Companies must then publish summaries of these evaluations so the public can see what they found.

After deploying an AI system, companies must check annually whether it's actually causing harm in the real world. If problems are discovered, they must take steps to fix them and report findings to the Federal Trade Commission. The law also requires companies to talk with communities that will be affected by the AI before using it. Companies must keep all evaluation records for at least 10 years and can't use AI in ways that create unfair discrimination, even if unintentional. The goal is to make sure AI systems used in consequential decisions are safe, fair, and don't secretly disadvantage certain groups of people.