State AI Laws Create a Patchwork for Consortia

State AI laws are the individual statutes and regulations that US states — rather than the US Congress — have enacted to govern the development, deployment and disclosure of artificial intelligence, and by mid-2026 more than 45 states have introduced such legislation with no unifying federal framework in place. For research consortia and shared-service research offices that span multiple states, this means the same AI-assisted grant-writing tool, chatbot, or automated screening system can be lawful in one member campus’s state and restricted or unlawful in another’s.

A state AI law is a statute enacted by an individual US state legislature — as distinct from federal legislation — that regulates how artificial intelligence systems are developed, deployed, disclosed, or audited within that state’s jurisdiction.

• What are state AI laws, and how many states have passed them?
• How do California, Colorado, and Texas AI laws compare?
• Is the federal government trying to preempt state AI laws?
• Why is the patchwork a problem for multi-campus research consortia?
• Answer-first Q&A
• Implications for shared-service research offices
• Outlook

What Are State AI Laws, and How Many States Have Passed Them?

State AI laws cover a wide range of obligations: algorithmic-discrimination audits, generative-AI content disclosure, “high-risk” system impact assessments, and rules on AI use in employment and consumer decisions. According to the National Conference of State Legislatures (NCSL), in the 2025 legislative session all 50 states, Puerto Rico, the Virgin Islands, and Washington, DC introduced AI-related legislation — a volume that state legislative trackers report continued to accelerate into 2026, with well over a thousand AI-related bills introduced nationwide.

No two states have adopted the same definitions, thresholds, or enforcement mechanisms. A tool classified as “high-risk automated decision-making” in one state’s statute may fall entirely outside another state’s scope, or be captured under a different label altogether.

How Do California, Colorado, and Texas AI Laws Compare?

Three states illustrate how far the approaches diverge. California took effect on 1 January 2026 with two distinct statutes: the Transparency in Frontier AI Act (SB 53), which imposes safety and transparency reporting duties on developers of large-scale frontier models, and the AI Transparency Act (SB 942), which requires disclosure when content is AI-generated. Colorado enacted the first comprehensive state AI statute, the Colorado AI Act (SB 24-205), which took effect in June 2026 and requires developers and deployers of high-risk AI systems to complete impact assessments and provide consumer disclosures; a subsequent amendment, SB 26-189, narrowed that scope to automated decision-making technology, with a revised effective date of 1 January 2027. Texas, by contrast, passed the Texas Responsible AI Governance Act (TRAIGA), which favours an industry self-governance framework over Colorado’s impact-assessment model.

Is the Federal Government Trying to Preempt State AI Laws?

Yes — but not yet through legislation that has passed Congress. On 11 December 2025, the White House signed an executive order, “Ensuring a National Policy Framework for Artificial Intelligence,” directing federal agencies to identify state AI laws that require models to alter their truthful outputs or that otherwise obstruct a national AI policy. The order instructs agencies to evaluate funding and litigation levers against such state statutes, but an executive order cannot itself repeal state law: only Congress or the courts can do that, and no comprehensive federal AI statute analogous to the EU AI Act has been enacted.

Until preemption legislation clears Congress — proposals exist but none has passed as of mid-2026 — state AI laws remain, in the words of one large law firm’s 2026 tracker, “the primary source of compliance obligations” for organisations operating in the United States.

Why Is the Patchwork a Problem for Multi-Campus Research Consortia?

Multi-state research consortia, shared-service research offices, and multi-site funded studies do not choose a single home jurisdiction the way a single-campus institution does. A consortium spanning California, Colorado, Texas, and a fourth state must reconcile at least three incompatible disclosure and assessment regimes simultaneously — and update that reconciliation as amendments such as Colorado’s SB 26-189 shift scope and effective dates mid-cycle.

This creates specific friction points for research administration:

• AI-assisted grant writing and proposal development may trigger content-disclosure duties in California but not in a partner state, complicating a single consortium-wide authorship and disclosure policy.
• Automated screening or scoring tools used in participant recruitment, peer review triage, or research-integrity checks can qualify as “high-risk automated decision-making” in Colorado while sitting outside any equivalent category in Texas.
• Shared IT and data infrastructure hosted in one state does not exempt a consortium from a partner campus’s home-state obligations when researchers in that state are end users of the system.
• Vendor contracts for AI writing, transcription, or analysis tools need jurisdiction-by-jurisdiction compliance riders rather than a single boilerplate clause.

Research offices increasingly need to disclose AI involvement in scholarly outputs regardless of state law, aligning with journal and funder expectations. Where AI tools contribute to drafting, structured contributor role taxonomies used in authorship disclosure — the model CASRAI originated in 2014 and which NISO now stewards as ANSI/NISO Z39.104-2022 — offer one consistent way to record human-versus-tool contribution that sits independently of any single state’s transparency statute.

Answer-First Q&A

How many US states have introduced AI legislation in 2026?

By the 2025 legislative session, all 50 states, Puerto Rico, the Virgin Islands, and Washington, DC had introduced AI-related legislation, according to the National Conference of State Legislatures. State legislative trackers report the volume of AI-related bills continued to climb through early 2026, spanning dozens of states with no sign of consolidation.

Does the federal executive order override state AI laws?

No. The December 2025 executive order directs federal agencies to identify and challenge state AI laws it views as obstructive, but an executive order cannot repeal state statute. Only an act of Congress or a binding court ruling can preempt state AI law, and no such federal AI statute has been enacted as of mid-2026.

What is the Colorado AI Act’s current status?

The original Colorado AI Act (SB 24-205) required impact assessments for high-risk AI systems. It was subsequently narrowed by SB 26-189 to focus specifically on automated decision-making technology, with a revised effective date of 1 January 2027, replacing the earlier June 2026 start date.

What should a multi-state research consortium put in its AI-use policy?

A consortium policy should map every member campus’s home-state AI statute, flag tools that trigger disclosure or impact-assessment duties in any one jurisdiction, and apply the strictest applicable standard consortium-wide rather than negotiating exceptions state by state.

Implications for Shared-Service Research Offices

Shared-service research offices — the units that run grants administration, research integrity, and compliance for several campuses at once — cannot rely on a single state’s AI statute as their reference point. The practical implication is that AI-use policy for a multi-campus consortium must be built to the strictest state standard among its members, then adjusted downward only where a specific campus’s home-state law is demonstrably more permissive and the consortium is willing to operate two policy tiers. Bodies such as NCURA and EARMA increasingly field member questions on exactly this cross-jurisdictional friction, reflecting how quickly the patchwork has become an operational, not just a legal, concern.

Consortium agreements and vendor contracts should each name which state-law regime governs AI tool use for that workflow, rather than assuming the lead institution’s state law applies uniformly to every partner.

Outlook

Absent a federal AI statute, the state-by-state pattern set by California, Colorado, and Texas is likely to keep expanding rather than converging in the near term. Consortia that govern to the strictest applicable state standard, and document AI contribution through structured, framework-neutral disclosure practices, will adapt faster as more states legislate and as amendments such as SB 26-189 continue to shift effective dates and scope. Treat this as a standing monitoring task, not a one-time policy update — state statutes are already being amended within their first year in force.