WHAT THIS BILL REGULATES · 8 REQUIREMENT TYPES
How Is This Bill Enforced
Verbatim statutory text on the left; plain-language analysis and a per-section checklist on the right. Numbered markers cross-link to the matching checklist row.
This act shall be known and may be cited as the "Responsible Deployment of AI SystemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) Act".
This section establishes the short title of the act as the Responsible Deployment of AI Systems Act. It creates no compliance obligations.
1 "AI systemAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1)" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation;
2 "DeployerDeployer"Deployer" means any public entity, private organization, or individual that implements AI systems for operational use.25 O.S. § 502(2)" means any public entity, private organization, or individual that implements AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) for operational use;
3(a) 1 "Risk classificationRisk classification"Risk classification" means one of the following: a. Unacceptable Risk: (1) means AI applications incompatible with social values and fundamental rights, (2) includes social scoring, manipulative AI targeting vulnerable groups, and real-time biometric identification systems or may include any of the following: (a) deployment of AI systems for discriminatory lending practices or biased law enforcement profiling, (b) unauthorized use of biometric analysis tools for surveillance purposes in public and private spaces, (c) integration of AI into systems with unregulated access to sensitive government databases, or (d) AI-driven misinformation campaigns targeting elections, public health, or emergency response systems, and (3) AI systems under this category are prohibited from development, deployment, or use, b. High Risk: (1) means AI systems with significant potential to impact safety, civil liberties, or fundamental rights, (2) includes AI in health care, critical infrastructure, law enforcement, financial services, and public welfare, and (3) AI systems under this category are subject to rigorous pre-deployment risk assessments, independent audits, strict operational requirements, and ongoing real-time monitoring including mandatory human oversight by qualified overseers, c. Limited Risk: (1) means AI systems that pose moderate risks, such as manipulation or deceit, but not classified as high risk, (2) includes chatbots and content creation tools such as deepfake generators, and (3) AI systems under this category require providers to ensure transparency, informing users when they interact with AI systems or consume AI-generated content and requires basic safeguards to prevent manipulation, and d. Minimal Risk: (1) means AI systems with little to no risk to users, (2) Examples include most consumer-facing AI applications such as virtual assistants, and (3) AI systems under this category require minimal oversight, with adherence to basic data protection standards.25 O.S. § 502(3)" means one of the following: a. Unacceptable Risk: (1) means AI applications incompatible with social values and fundamental rights, (2) includes social scoring, manipulative AI targeting vulnerable groups, and real-time biometric identification systems or may include any of the following: (a) deployment of AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) for discriminatory lending practices or biased law enforcement profiling, (b) unauthorized use of biometric analysis tools for surveillance purposes in public and private spaces, (c) integration of AI into systems with unregulated access to sensitive government databases, or (d) AI-driven misinformation campaigns targeting elections, public health, or emergency response systems, and (3) AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) under this category are prohibited from development, deployment, or use,
3(b)–(d) b. High Risk: (1) means AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) with significant potential to impact safety, civil liberties, or fundamental rights, (2) includes AI in health care, critical infrastructure, law enforcement, financial services, and public welfare, and (3) AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) under this category are subject to rigorous pre-deployment risk assessments, independent audits, strict operational requirements, and ongoing real-time monitoring including mandatory human oversight by qualified overseers, c. Limited Risk: (1) means AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) that pose moderate risks, such as manipulation or deceit, but not classified as high risk, (2) includes chatbots and content creation tools such as deepfake generators, and (3) AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) under this category require providers to ensure transparency, informing users when they interact with AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) or consume AI-generated content and requires basic safeguards to prevent manipulation, and d. Minimal Risk: (1) means AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) with little to no risk to users, (2) Examples include most consumer-facing AI applications such as virtual assistants, and (3) AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) under this category require minimal oversight, with adherence to basic data protection standards;
4 "Qualified human overseerQualified human overseer"Qualified human overseer" means a trained individual responsible for monitoring and validating high-risk AI system outputs, with authority to amend or override them when necessary.25 O.S. § 502(4)" means a trained individual responsible for monitoring and validating high-risk AI systemAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) outputs, with authority to amend or override them when necessary.
This section defines the four key terms used throughout the act: AI system, Deployer, Risk classification (with its four tiers — unacceptable, high, limited, and minimal), and Qualified human overseer. Notably, the risk-classification definition itself embeds operative requirements: unacceptable-risk systems are prohibited, high-risk systems require rigorous assessments and human oversight, limited-risk systems require transparency, and minimal-risk systems require basic data protection. The prohibition on unacceptable-risk AI systems is the most consequential provision embedded here.
A This act is applicable to all public sector agencies and private entities deploying artificial intelligence (AI) systems affecting individuals within this state.
B All AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) shall comply with existing laws, including: 1. Federal and state civil rights protections; 2. Data protection laws and state privacy laws; and 3. Consumer protection laws and regulations.
This section establishes the act's jurisdictional scope — all public sector agencies and private entities deploying AI systems that affect individuals within Oklahoma — and requires all AI systems to comply with existing federal and state civil rights, data protection, privacy, and consumer protection laws. The existing-law compliance mandate does not create new obligations beyond what those other statutes already require, but it confirms that AI deployment does not exempt entities from preexisting legal frameworks.
A 2 DeployersDeployer"Deployer" means any public entity, private organization, or individual that implements AI systems for operational use.25 O.S. § 502(2) must classify artificial intelligence (AI) systems into one of four risk categories outlined in Section 2 of this act before deployment.
B 3 DeployersDeployer"Deployer" means any public entity, private organization, or individual that implements AI systems for operational use.25 O.S. § 502(2) shall conduct assessments of AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) to identify: 1. Potential biases in training data; 2. Risks to safety, civil liberties, and fundamental right; and 3. Mitigation strategies for identified risks.
C 4 High-risk AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) must undergo ongoing performance evaluations, with documentation of findings and actions taken to address deficiencies.
D 5 Qualified overseers must validate AI outputs before they are enacted and retain authority to override system recommendations.
E 6 DeployersDeployer"Deployer" means any public entity, private organization, or individual that implements AI systems for operational use.25 O.S. § 502(2) shall notify affected individuals when high-risk AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) influence decisions about the individuals and provide avenues for appeal or human review.
Section 504 imposes the core operational compliance obligations on deployers. It requires pre-deployment risk classification of every AI system into one of the four tiers defined in § 502, pre-deployment bias and safety assessments, ongoing performance evaluations with documented findings for high-risk systems, mandatory human oversight with override authority via qualified overseers, and individual notification with appeal rights when high-risk AI influences decisions about individuals. These obligations collectively create a comprehensive pre- and post-deployment compliance regime for high-risk AI.
A 7 All deployersDeployer"Deployer" means any public entity, private organization, or individual that implements AI systems for operational use.25 O.S. § 502(2) shall establish governance groups to oversee the classification, deployment, and monitoring of high-risk artificial intelligence (AI) systems.
B 8 DeployersDeployer"Deployer" means any public entity, private organization, or individual that implements AI systems for operational use.25 O.S. § 502(2) shall maintain the following: 1. Updated inventories of deployed AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1); 2. Documentation on system design, intended use, and training data; and 3. Records of audits, risk assessments, and oversight activities.
C 9 High-risk AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) must undergo annual independent audits to ensure compliance with legal and ethical standards.
D 10 DeployersDeployer"Deployer" means any public entity, private organization, or individual that implements AI systems for operational use.25 O.S. § 502(2) shall implement protocols to address identified deficiencies, including suspension or recalibration of systems.
Section 505 requires all deployers to establish internal governance groups for the classification, deployment, and monitoring of high-risk AI systems. Deployers must maintain updated inventories of deployed AI systems, documentation on system design, intended use, and training data, and records of audits, risk assessments, and oversight activities. High-risk AI systems must undergo annual independent audits for legal and ethical compliance. Additionally, deployers must implement remediation protocols — including suspension or recalibration — to address identified deficiencies.
A 11 DeployersDeployer"Deployer" means any public entity, private organization, or individual that implements AI systems for operational use.25 O.S. § 502(2) shall provide accessible channels for public and stakeholder feedback on the operation and impact of artificial intelligence (AI) systems.
B 12 Aggregated feedback data must be included in an annual performance report made available by January 1 of each year to the Artificial Intelligence Council (AI Council), created in Section 8 of this act.
C The AI Council will analyze feedback and make annual recommendations to inform updates to state and industry-specific artificial intelligence and machine-learning regulations.
Section 506 creates a public feedback and reporting loop. Deployers must provide accessible channels for public and stakeholder feedback on AI system operation and impact. Aggregated feedback data must be compiled into an annual performance report submitted to the AI Council by January 1 each year. The AI Council is then directed to analyze the feedback and make annual recommendations to inform updates to state and industry-specific AI and machine-learning regulations.
A Violations of this act may result in the Artificial Intelligence Council (AI Council) issuing any of the following: 1. Financial penalties proportional to the severity of the infraction; 2. Suspension or prohibition of noncompliant systems; and 3. Mandatory external oversight until compliance is restored.
B 13 DeployersDeployer"Deployer" means any public entity, private organization, or individual that implements AI systems for operational use.25 O.S. § 502(2) shall publicly disclose noncompliance incidents and the corrective actions taken.
Section 507 establishes the enforcement regime. The AI Council may issue financial penalties proportional to infraction severity, suspend or prohibit noncompliant systems, and mandate external oversight until compliance is restored. Separately, deployers are required to publicly disclose noncompliance incidents and the corrective actions taken — a standalone transparency obligation distinct from the penalty regime.
A There is hereby established the Artificial Intelligence Council (AI Council) which shall oversee the regulation, deployment, and ethical use of AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) within this state.
B The AI Council shall be responsible for: 1. Overseeing the AI Regulatory Sandbox Program to foster innovation while maintaining safety and accountability; 2. Manage the AI Workforce Development Program, including administering grants to small businesses and educational institutions, with a focus on underserved communities; 3. Develop and update guidelines for AI systemAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) classification, risk assessments, and ethical standards; 4. Facilitate public and stakeholder engagement on AI policies and regulations; and 5. Provide annual reports to the Legislature on the status of AI deployment, innovation, and workforce initiatives.
C The AI Council shall include nine (9) members: 1. Three appointed by the Governor; 2. Three appointed by the Speaker of the Oklahoma House of Representatives; and 3. Three appointed by the President Tempore of the Oklahoma State Senate.
D Selected members shall come from one or more of the following backgrounds: 1. Experts in artificial intelligence ethics, law, and technology; 2. Representatives from public agencies and private industry; and 3. Advocates for civil rights and underserved communities.
Section 508 establishes the Artificial Intelligence Council as the state's primary AI regulatory body, consisting of nine members appointed by the Governor, the Speaker of the House, and the President Pro Tempore of the Senate. The AI Council's responsibilities include overseeing the AI Regulatory Sandbox Program, managing the AI Workforce Development Program, developing classification and ethical guidelines, facilitating public engagement, and providing annual reports to the Legislature. Members must have expertise in AI ethics, law, technology, public or private sector operations, or civil rights advocacy. This section creates institutional infrastructure rather than compliance obligations on deployers.
A–B A. The Artificial Intelligence Council (AI Council) shall develop the AI Regulatory Sandbox Program. B. The AI Regulatory Sandbox Program shall provide a controlled environment for deployersDeployer"Deployer" means any public entity, private organization, or individual that implements AI systems for operational use.25 O.S. § 502(2) to test innovative AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) while ensuring compliance with ethical and safety standards.
C The AI Council shall: 1. Approve or deny applications for participation in the sandbox; 2. Set limits and safeguards for testing activities; and 3. Monitor and evaluate sandbox outcomes for compliance and potential risks.
D Applicants must demonstrate: 1. The potential for societal benefit or economic development; 2. Compliance with data privacy and security requirements; and 3. Alignment with ethical principles set by the AI Council.
E 14 Participants shall submit periodic progress reports and a final evaluation detailing the system's performance, risks identified, and mitigation measures taken.
Section 509 directs the AI Council to develop an AI Regulatory Sandbox Program providing a controlled testing environment for deployers to pilot innovative AI systems while maintaining ethical and safety compliance. The AI Council approves or denies applications, sets testing limits and safeguards, and monitors outcomes. Applicants must demonstrate societal benefit or economic development potential, data privacy compliance, and alignment with AI Council ethical principles. Participants must submit periodic progress reports and a final evaluation. This section primarily creates obligations on the AI Council and sandbox participants rather than on deployers generally.
A The Artificial Intelligence Council (AI Council) shall develop the Artificial Intelligence Workforce Development Program (AI Program), the purpose of which shall be to enhance AI-related workforce capabilities by focusing on small businesses and underserved communities, particularly rural and low-income areas.
B The AI Council shall: 1. Develop grant criteria for small businesses and educational institutions; and 2. Monitor program outcomes and report annually to the Legislature.
C Eligible applicants may include: 1. Small businesses integrating AI systemsAI system"AI system" means an artificial intelligence or machine learning-based or algorithmic technology designed to perform tasks that typically require human intelligence, including decisionmaking, prediction, or recommendation.25 O.S. § 502(1) to improve operations; and 2. Educational institutions offering AI-related training or certification programs targeting underserved populations.
D The program shall prioritize: 1. Rural outreach programs to introduce AI literacy and training; 2. Partnerships with local schools to create STEM pathways focused on AI; and 3. Affordable training opportunities for individuals from low-income backgrounds.
Section 510 directs the AI Council to develop an AI Workforce Development Program focused on enhancing AI-related capabilities for small businesses and underserved communities. The AI Council develops grant criteria, monitors program outcomes, and reports annually to the Legislature. Eligible applicants include small businesses integrating AI and educational institutions offering AI training targeting underserved populations. The program prioritizes rural outreach, school STEM partnerships, and affordable training for low-income individuals. This section creates no compliance obligations on deployers.
The Artificial Intelligence Council (AI Council) shall have the authority to establish rules to enforce this act.
Section 511 grants the AI Council rulemaking authority to enforce the act. This is an enabling provision that creates no direct compliance obligation on deployers but authorizes the AI Council to promulgate implementing rules.
This act shall become effective November 1, 2025.
The act becomes effective November 1, 2025.