State lawmakers across the country enacted 115 artificial intelligence laws in 2026 — a nearly 58 percent surge from the prior year — according to a comprehensive full-year report released today by the Transparency Coalition, a Seattle-based nonprofit that tracks AI policy across all 50 states. The 2026 State AI Legislation Report covers laws passed or enacted in 30 states between January 1 and October 15, up from 73 laws across 27 states in the organization's 2025 annual count.

The numbers reflect a structural shift in where AI governance actually happens in the United States. Congress has not enacted a single federal AI law despite more than 150 federal AI bills introduced across the past two sessions. The White House has instead pushed for federal preemption of state AI rules — an effort that collapsed in the Senate on a 99-1 vote in July 2025. A "Great American AI Act" preemption discussion draft, floated in June 2026 by a bipartisan House pair, remains stalled. The result, by default, is that the states are writing American AI law.

"State lawmakers in both parties have stepped up to the challenge of regulating AI for the good of our communities," said Rob Eleveld, CEO of the Transparency Coalition. "What stands out this year is the seriousness with which lawmakers across the country are tackling this challenge."

Children and Chatbots: Fourteen States Move Against AI-Powered Harm

The single most active legislative category in 2026 was chatbot safety, with 14 states advancing laws addressing risks including self-harm, suicidal ideation, sexually explicit content, and emotionally manipulative engagement tactics designed to keep children interacting with AI-powered chat products. California, Oregon, and Washington enacted particularly expansive protections.

The legislative urgency is grounded in documented cases. In 2023, a 13-year-old in Colorado died by suicide following interactions with a Character.ai chatbot. In May 2026, Pennsylvania sued Character Technologies after a chatbot named "Emilie" presented itself to roughly 45,500 users as a licensed therapist with fabricated credentials. A 2025 study by Sentio University found that none of six major AI chatbots fully met clinician safety standards when users raised suicide or mental health crises. The AMA has urged Congress to prohibit chatbots from diagnosing or treating mental health conditions.

"Too many of the rising AI corporations are revealing themselves to be as ruthless and predatory — if not more so — than their social media predecessors," Eleveld wrote in the foreword to the organization's mid-year 2026 report.

Related to chatbot safety, seven states — California, Colorado, Connecticut, Idaho, Illinois, Minnesota, and South Carolina — also adopted measures targeting addictive or compulsive design features in digital products used by children and teenagers, including algorithmic feeds, autoplay mechanics, and engagement reward systems.

Healthcare: Eight States Say a Human Must Stay in the Loop

A second major cluster of 2026 legislation addressed AI's expanding role in healthcare, and specifically in decisions with direct consequences for patient care.

Eight states — Alabama, Colorado, Georgia, Illinois, Indiana, Iowa, Utah, and Washington — enacted laws limiting the ability of health insurers to rely solely on AI or automated systems when making adverse authorization or coverage decisions. The laws vary in scope but share a core requirement: a human reviewer must remain in the decision loop when an AI system might deny or restrict patient care. If your insurer in one of these eight states denies a claim, they are now legally required to have had a person involved in that decision — not just an algorithm.

Five states went further on mental health specifically. Colorado, Maine, Rhode Island, Tennessee, and Vermont prohibited or restricted the offering of AI-delivered mental health therapy, relying on existing state authority over licensed healthcare professions. The Transparency Coalition noted that lawmakers may extend similar professional-licensing approaches to other regulated fields in 2027 as states increasingly treat licensing authority as a brake on AI substitution for licensed practitioners.

Illinois Sets the Frontier: What the Nation's First Mandatory AI Audit Actually Requires

The most consequential single piece of legislation in the report is Illinois Senate Bill 315, the Artificial Intelligence Safety Measures Act, signed by Governor J.B. Pritzker on July 6, 2026.

Illinois became the third state — after California and New York — to enact comprehensive frontier AI legislation. But it went further than either predecessor on one critical point: Illinois is the first jurisdiction in the United States to mandate independent third-party audits of the largest AI developers, according to Morrison Foerster's analysis of the law.

What counts as a frontier model under the law: A general-purpose AI system trained using more than 10²⁶ floating-point operations (FLOP). For context, the EU AI Act sets its analogous threshold at roughly 10²⁵ FLOP — Illinois's bar is ten times higher, deliberately calibrated to cover only the largest models. At 10²⁶ FLOP, the companies in scope are a short list: the class of models that includes GPT-4 and successors, Gemini Ultra, Claude 3 Opus-class systems, and Meta's largest Llama models.

Who must comply: Companies that train frontier models above that threshold and exceed $500 million in annual gross revenue (counting all affiliated entities in the preceding calendar year) must retain a qualified, independent third party to conduct annual compliance audits starting in 2028.

What auditors must be: The law requires demonstrated competence in frontier model safety and prohibits any financial interest in the developer being audited. Audit summaries must be published within 30 days of completion and transmitted to the Illinois Emergency Management Agency and the state Attorney General. Unredacted versions are retained by the developer for the duration of deployment plus five additional years.

What else the law requires: All frontier developers — regardless of revenue — must file a public disclosure statement with the Agency before developing or deploying in Illinois, beginning January 1, 2027. They must publish machine-readable transparency reports covering release dates, supported languages, output types, intended uses, and deployment restrictions. Critical safety incidents must be reported to the Agency within 72 hours, or within 24 hours if death or serious injury is imminent. Companies that fail to meet disclosure obligations face civil penalties of up to $1,000 per day; violations of the broader act carry AG-imposed penalties up to $1 million for a first offense and up to $3 million for each subsequent violation.

Notably, both OpenAI and Anthropic publicly supported Illinois SB 315 — a split from the typical industry stance toward new regulatory mandates. The Computer and Communications Industry Association formally requested a veto on innovation grounds. Governor Pritzker signed it in part because, he said, the federal government was unwilling to intervene.

Does the Auditing Industry Actually Exist? A Bootstrapping Problem

Illinois's audit mandate presupposes an industry of qualified, independent frontier AI auditors — a profession that does not yet exist at the scale the law will require. The bill specifies competence in "frontier model safety" with no financial interest in the developer. No credentialed, independent auditing framework for frontier AI models currently operates at scale anywhere in the world.

California has begun addressing this gap directly. The state enacted separate 2026 measures establishing early standards and infrastructure for independent AI auditing organizations — building supply-side capacity for the audit ecosystem that Illinois's demand-side mandate will require. The Transparency Coalition's 2027 forecast identifies independent AI auditing expansion as among the developments most likely to advance in the coming year, including the possibility that more states will follow Illinois and California in establishing certification, licensing, and standards for third-party auditors.

Legal analysts have also noted that Illinois SB 315 may face preemption challenges from the federal AI Litigation Task Force before its first audit cycle begins in 2028.

What Frontier AI Law Is Being Used to Fight

Not everyone agrees that a state-by-state approach to frontier AI governance is wise, even among those who want stronger AI accountability.

"It's mostly bad," said Anton Dahbura, executive director of the Johns Hopkins Institute for Assured Autonomy. "AI crosses state lines. Inconsistent rules hurt U.S. competitiveness and create compliance nightmares for companies." Dahbura argues the fragmented patchwork creates uncertainty that may slow innovation without producing coherent safety outcomes, and that Congress must act.

The critique carries weight from a federalism standpoint. The Trump administration's DOJ AI Litigation Task Force, established January 10, 2026 under a December 2025 executive order challenging state AI laws, has been actively challenging state AI laws in federal court on Dormant Commerce Clause grounds — the legal theory that state laws unconstitutionally burden interstate commerce when they regulate products that inherently cross state lines. The White House also conditioned federal broadband funding on states' willingness to review or repeal AI regulations it considered burdensome.

The bipartisan nature of state-level support for AI regulation — reflected in the 99-1 Senate vote that rejected preemption in July 2025 — suggests the federal challenge will face significant political headwinds even if it has legal traction.

Transparency, Deepfakes, and Education: The Rest of the 2026 Agenda

Beyond the headline categories, the Transparency Coalition's full-year report details several other threads of legislative activity that illustrate how broadly state lawmakers are engaging with AI.

On disclosure and transparency, states continued to require clearer notifications when AI is involved in advertising, customer service, campaign communications, and chatbot interactions. Washington enacted a law requiring covered AI developers to embed provenance information directly into AI-generated or AI-altered media — an approach intended to make synthetic content identifiable at the file level, not just through platform-level labels.

On deepfakes, states including California, Hawaii, Kansas, Louisiana, Maryland, Tennessee, Utah, and Washington strengthened protections against harmful or deceptive synthetic media, building on frameworks established in earlier legislative sessions.

On education, states including California, Illinois, Maryland, Oklahoma, Tennessee, Utah, and Virginia enacted measures addressing AI procurement in schools, classroom use guidelines, and the role of human teachers.

California alone enacted nearly 30 AI-related measures during the 2026 legislative year — a volume that, by itself, would have exceeded all state AI legislation from 2024.

What to Expect in 2027

The Transparency Coalition's report concludes with a 2027 forecast that identifies several issues likely to move to the foreground.

AI product liability is emerging as a distinct focus as lawmakers examine how traditional product-liability principles should apply to AI products capable of causing foreseeable harm — the Pennsylvania chatbot lawsuit being an early example of the litigation landscape that will pressure legislatures to act.

Stronger chatbot safeguards are expected as first-generation laws prove inconsistent in their requirements and documented harms continue to accumulate.

Independent AI auditing is expected to expand, with more states potentially following Illinois and California in establishing certification, licensing, and standards for third-party auditors — creating what the Transparency Coalition calls "a new professional services market from regulatory necessity."

Public procurement is on the horizon as a safety lever, with state agencies and public school systems drawing new legislative attention as major technology buyers who could require higher AI safety and transparency standards as a condition of purchase.

The full 2026 State AI Legislation Report, including a state-by-state guide to all 115 laws, is available at the Transparency Coalition's website.

Frequently Asked Questions

What does Illinois's new AI audit law actually require, and does it apply to me?

Illinois SB 315, the Artificial Intelligence Safety Measures Act, requires annual independent third-party compliance audits for companies that train "frontier" AI models — defined as general-purpose AI systems trained using more than 10²⁶ floating-point operations — and that also exceed $500 million in annual gross revenue. That threshold covers only a handful of companies globally (the developers of the largest current commercial AI systems). If you are an individual user or a small developer, the audit mandate does not apply to you directly. However, the law also requires all frontier AI developers — regardless of revenue — to publish machine-readable transparency reports and file public disclosure statements with the state beginning January 1, 2027. Audit obligations for large frontier developers begin January 1, 2028.

If my insurer used AI to deny a healthcare claim, is that now illegal?

In eight states — Alabama, Colorado, Georgia, Illinois, Indiana, Iowa, Utah, and Washington — health insurers are now prohibited from relying solely on AI or automated systems to make adverse coverage or authorization decisions. If you live in one of these states and received an AI-generated coverage denial with no human reviewer in the process, your insurer may have violated state law. You can contact your state insurance commissioner's office to file a complaint. If you live outside those eight states, no equivalent protection currently exists at the federal level.

Can an AI chatbot legally pose as a therapist?

In at least five states — Colorado, Maine, Rhode Island, Tennessee, and Vermont — AI products that provide mental health therapy services face legal restrictions under professional licensing laws. Pennsylvania has sued Character Technologies in court over a chatbot that allegedly presented as a licensed therapist to tens of thousands of users. Outside those states, no federal law currently bars AI chatbots from adopting therapeutic personas. The regulatory picture is changing rapidly: the American Medical Association has called on Congress to prohibit chatbots from diagnosing or treating mental health conditions, and additional state legislation is forecast for 2027.

Who will actually conduct these mandatory frontier AI audits — and does that profession exist yet?

This is the bootstrapping problem at the heart of Illinois's landmark law. The bill requires auditors with "demonstrated competence in frontier model safety" and no financial interest in the developer — a professional standard that does not yet map to an established, credentialed field. No large-scale, accredited frontier AI auditing industry currently exists. California has begun building supply-side infrastructure — standards and early certification frameworks for independent AI auditing organizations — that would provide auditors capable of conducting the work Illinois's law will demand beginning in 2028. The Transparency Coalition expects independent AI auditing to expand significantly in 2027 as additional states consider similar requirements.