H.R. 5511: Algorithmic Accountability Act of 2025

The Algorithmic Accountability Act of 2025 would force large companies to test the AI systems they use for what the bill calls "critical decisions" — jobs, housing, loans, healthcare, education, utilities, family planning, and legal services. Companies would have to check those systems for bias, privacy risks, and harm both before and after deploying them, then file reports with the Federal Trade Commission.

The rules would hit two tiers. Tier 1 catches any company with more than $50 million in average annual gross receipts, more than $250 million in equity value over a 3-year period, or data on more than 1 million consumers, households, or devices. Tier 2 reaches smaller firms — those over $5 million in receipts or $25 million in equity — if they build algorithms for Tier 1 clients. Both thresholds rise annually with the Consumer Price Index after the first year.

The FTC would have 2 years to write the rules, and the rules would take effect 2 years after that. Once live, covered companies would have to run impact assessments examining differential performance by race, color, sex, gender, age, disability, religion, family status, socioeconomic status, and veteran status. They'd also have to evaluate whether consumers get notice, opt-out options, transparency, explainability, and a way to contest decisions. Documentation would have to be kept for 3 years longer than the algorithm stays in use.

To enforce all this, the bill creates a new Bureau of Technology inside the FTC, led by a Chief Technologist with at least 50 staff. The Bureau of Consumer Protection gets 25 additional enforcement personnel. A public FTC repository — searchable, sortable, and downloadable — would publish summary reports within 180 days of the rules taking effect, updated quarterly. State attorneys general could also sue companies that violate the rules. The bill doesn't preempt stricter state, tribal, city, or local laws.