Sectoral AI Governance Act of 2026

Key Points

• This bill gives every federal agency with enforcement power the authority to create new rules targeting AI and algorithmic systems that are likely to help cause violations of existing federal laws. Agencies could regulate hiring tools, loan software, health algorithms, and more.
• The bill defines 'algorithmic decision-making system' very broadly to include any computational process that makes or meaningfully influences decisions, covering AI, machine learning, and statistical models. It specifically excludes basic tech like web hosting and data storage.
• Before writing any new AI rule, agencies must publish an advance notice to get public feedback and consult with the Office of Information and Regulatory Affairs, the Office of Science and Technology Policy, and NIST to check for conflicts with other agencies and get technical guidance.
• Agencies must review their AI rules within 3 years and at least every 5 years after that. If technology has changed or the rule no longer fits, the agency must update or scrap it, building in a sunset-style check to prevent outdated regulations.
• The bill requires biennial reports to Congress from every agency that used this authority, as well as from agencies that identified harmful AI uses but chose not to write rules. Reports must cover enforcement actions, staffing, and any discriminatory effects found.
• States can still pass their own AI regulations. Federal rules only override state laws when there is a direct conflict, preserving a patchwork approach where states can go further than the federal government.