Technology; artificial intelligence; state agencies; prohibited uses; permitted uses; Office of Management and Enterprise Services; effective date.
HB 3545 creates a statewide framework governing how Oklahoma state agencies may use artificial intelligence. It defines key terms such as artificial intelligence systems, generative AI, deepfakes, and “high-risk AI decisions,” and then sets out both prohibited and permitted uses. The bill bars agencies from using AI for cognitive behavioral manipulation, discriminatory classification based on behavior or socioeconomic status, real-time remote biometric surveillance in public spaces except under narrow warrant, court order, or exigent-circumstance exceptions, and deceptive or malicious deepfakes.
The bill also allows AI use in state government, but only with safeguards. High-risk AI decisions must be reviewed and approved by a trained human decision-maker before taking effect, AI-generated material that has not been reviewed by a human must be disclosed as AI-generated, and users must be informed when they are interacting with AI. Agencies must review their systems within nine months after the effective date, remove prohibited systems, update inconsistent procedures, and ensure future AI deployments comply with the act. OMES must compile and publish annual reports identifying agency AI systems, prohibited systems removed, and compliance procedures.
The bill would add a new section to Title 75A of the Oklahoma Statutes and directly regulate AI procurement and use across all state agencies, including the legislative and judicial branches, with limited exceptions for research systems at state-funded higher education institutions and common consumer systems. It would require agencies to audit existing AI tools, revise internal procedures, and maintain ongoing compliance reporting to the Office of Management and Enterprise Services. OMES would also gain a continuing oversight and reporting role, including annual public reporting to state leaders and posting reports online.
The available committee and floor history shows strong support and no recorded opposition. The bill passed the House Government Modernization and Technology Committee 7-0, the House Commerce and Economic Development Oversight Committee 14-0, and third reading in the House 93-0. The committee transcript reflects routine adoption of a proposed committee substitute and brief procedural remarks, suggesting the measure was broadly accepted and not heavily contested in the recorded proceedings.
The main policy tensions in the bill concern privacy, surveillance, and administrative discretion. The most notable restrictions are the limits on real-time biometric identification and facial recognition in public spaces, which are allowed only with a warrant, court order, or narrowly defined emergency circumstances. Another potential point of concern is the human-review requirement for high-risk AI decisions, which could be viewed as necessary oversight by supporters but as an operational constraint by agencies. The bill also addresses deepfakes and disclosure of AI-generated content, reflecting concern about deception and transparency in government communications.