AI Frontier Institute

Technology; artificial intelligence; state agencies; prohibited uses; permitted uses; Office of Management and Enterprise Services; effective date.

View primary source →

Engrossed Oklahoma House Bill 3545 has passed the House (96-0) and is now in the Senate, so it is not law. If enacted, it would add Section 200 of Title 75A of the Oklahoma Statutes to regulate how state agencies use artificial intelligence. "State agency" is defined broadly and includes the legislative and judicial branches and law enforcement agencies. The bill would apply to all state agency computer systems. It excepts systems used in research by state-funded higher education institutions and installed consumer systems in common personal use, such as phone face-unlock. Section 1(C) would prohibit four uses of AI by state agencies. The first is cognitive behavioral manipulation of people or groups. The second is classifying people in ways that result in unlawful discrimination. The third is real-time remote biometric identification, such as facial recognition, for surveillance in public spaces. That is allowed only with a valid warrant or court order, or in exigent circumstances involving an imminent threat to life or serious bodily injury. Exigent use must be limited in scope and duration and followed by documentation and supervisory review. The fourth is deepfakes used for deceptive or malicious purposes. Section 1(D) would allow other uses with conditions. A "high-risk AI decision" is an automated decision that is the primary basis for a final agency action affecting legal rights, benefits eligibility, civil penalties, or legal obligations, with a substantial impact and no meaningful human review. No such decision may take effect unless a trained human with decision-making authority reviews and approves it. Generative AI material that a human has not reviewed must carry an AI-generated disclosure. Users interacting with an AI system must be told they are doing so. Under Section 1(E), agencies would have nine months after the effective date to review their AI use, remove prohibited systems, revise inconsistent procedures, and report compliance to the Office of Management and Enterprise Services (OMES). Under Section 1(F), newly deployed AI systems and procedures must comply. Under Section 1(G), starting December 1, 2026 and every year after, OMES would report to the Governor, the House Speaker, and the Senate President Pro Tempore on AI systems, which were prohibited and removed, which are allowed, and the compliance procedures. OMES would also post the reports on its website. Under Section 2, the act would take effect November 1, 2026.

← Back to the tracker