H5123 creates the “Artificial Intelligence Accountability Act” and places new oversight requirements on Rhode Island state government’s use of artificial intelligence. The bill directs the Department of Administration to compile an annual inventory of all AI systems used by state agencies, beginning by December 31, 2026, and to publish that inventory on the state’s open data portal. The inventory must identify each system, its vendor, its general uses, whether it was used to make or support decisions, and whether an impact assessment was completed before implementation.
The bill also requires the Department of Administration to develop statewide policies and procedures by February 1, 2027 governing the procurement, implementation, and ongoing review of AI systems used by state agencies. Those policies must be designed to prevent unlawful discrimination and unlawful disparate impacts on protected and other specified groups, and state agencies would be prohibited from implementing AI systems unless an impact assessment has been completed or if the agency head determines the system would cause prohibited discrimination or disparate impact. In addition, the bill establishes a permanent artificial intelligence commission within the Department of Administration to study AI use in state government, examine transparency, auditability, due process, data security, intellectual property issues, and disparate impacts, and recommend best practices and policy changes.
The bill would affect Title 42 of the Rhode Island General Laws by adding a new chapter on AI accountability. It would impose new duties on the Department of Administration, require state agencies to conduct pre-implementation assessments, and create ongoing monitoring and reporting obligations. The commission would include executive branch officials, judicial and legal representatives, and academic experts, and would be required to submit annual public reports to the governor and legislative leaders with findings, recommendations, cost estimates, and implementation timelines.
The overall sentiment reflected in the bill materials is cautious and oversight-oriented rather than oppositional or celebratory. The proposal frames AI as a tool that can affect human welfare and legal rights, and it emphasizes transparency, accountability, and protection against discriminatory outcomes. Because there are no committee transcripts or recorded votes provided, there is no documented public debate in the supplied materials, but the structure of the bill suggests a policy interest in regulating government AI use before broader deployment.
The main points of contention likely concern the scope of state oversight, the administrative burden of inventories and impact assessments, and the potential limits on agency flexibility in adopting AI tools. The bill also raises issues around data access, vendor confidentiality, trade secrets, and whether the state can meaningfully audit complex AI systems while protecting proprietary information. Another likely area of concern is how broadly the bill defines disparate impact and which characteristics or outcomes should trigger restrictions on AI use.
The bill would add a new chapter to Title 42 governing artificial intelligence in state government, requiring statewide inventory, policy development, impact assessments, and ongoing monitoring of AI systems used by state agencies. It would also create a permanent AI commission within the Department of Administration and require annual public reporting, thereby expanding administrative oversight and transparency obligations for state agencies, vendors, and officials involved in AI procurement and use.
The bill’s tone and structure indicate generally favorable sentiment toward stronger oversight of government AI, with a focus on accountability, transparency, and anti-discrimination safeguards. No votes or committee testimony are provided, so there is no recorded opposition or support in the supplied materials, but the proposal appears designed as a precautionary regulatory framework rather than a deregulatory or innovation-first measure.
Likely points of contention include whether the Department of Administration should have broad authority to set AI policy for all state agencies, how costly and burdensome the required inventories and impact assessments will be, and whether the state can effectively evaluate AI systems without access to proprietary vendor information. The bill’s anti-discrimination and disparate-impact standards may also draw debate over how they are defined, enforced, and balanced against agency efficiency, innovation, and procurement flexibility.