AI Frontier Institute

Providing consumer protections for artificial intelligence systems.

View primary source →

Substitute Senate Bill 6284 has not been enacted. The Senate Environment, Energy & Technology Committee passed it as a substitute and referred it to Ways & Means, which held a hearing on 2026-02-06. If enacted, it would set up a risk-based framework against algorithmic discrimination by "high-risk" AI systems. These are systems meant to make, or be a substantial factor in making, a "consequential decision" without meaningful human consideration. Consequential decisions cover parole and probation, education, employment, financial or lending services, essential government services, health care, housing, insurance and legal services (Sec. 2). Deployers: Beginning July 1, 2027, deployers would have to use industry-standard means to protect consumers from known or reasonably foreseeable algorithmic discrimination (Sec. 3). Compliance with the chapter would create a rebuttable presumption of reasonable care in attorney general enforcement. Deployers would have to review each system at least annually and report discovered discrimination to the attorney general within 90 days (Sec. 3). They would also have to keep a risk management policy and program (Sec. 4) and complete impact assessments, including after substantial modifications. They must keep these records for at least three years (Sec. 6). Developers: Beginning July 1, 2027, developers would have to keep a risk management policy and program, with an exemption for developers with fewer than 50 full-time equivalent employees (Sec. 5). Small-deployer exemption: Some deployers with fewer than 50 employees that meet certain conditions are exempt from Secs. 4 and 6 (Sec. 7). Consumer notice: Beginning July 1, 2026, deployers would have to notify consumers before a consequential decision is made. They would also have to give a statement of the system's purpose, the deployer's contact information, and a plain-language description (Sec. 8). Exemptions: Sec. 9 exempts or carves out several categories. These include federally approved or certified systems, certain federal contract work, HIPAA covered entities for certain uses, insurers regulated by the insurance commissioner, and financial institutions. Whoever claims an exemption bears the burden of showing it applies. Enforcement: Only the attorney general may enforce, under the Consumer Protection Act. The bill creates no private right of action. The attorney general must give 45 days' written notice, and a first violation can be cured within 60 days (Sec. 10). Government agencies: An agency that offers an AI system meant to interact with consumers must clearly disclose that the consumer is interacting with AI, without dark patterns (Sec. 11). AI task force and workplace group: The bill amends 2024 c 163 s 2 (Sec. 14). The task force would no longer depend on appropriations, its final report would move to July 1, 2027, and its expiration to June 30, 2028. It also creates an AI workplace advisory group (Sec. 15). The advisory group would develop guiding principles for workplace AI. Its interim report is due December 1, 2026, and its final report March 1, 2027. The group expires June 30, 2028.

← Back to the tracker