State AI Laws in 2026: SB 53, SB 942, and Preemption

Updated October 2026

State AI laws went from a watch-list item to real work on your desk. Two California statutes now bind, Texas and Illinois switched on statutes of their own on New Year's Day, and a Department of Justice task force is in court trying to unwind Colorado's.

David Morris, General Counsel of Snyk at the time of recording and now Chief Legal Officer at Darktrace, saw the board pressure coming on CZ and Friends, GC AI's podcast with legal leaders:

You have to, you can't not engage with them because it's not comfortable or it's not how your CEO or CFO or leaders want to spend their time. They don't need to spend a lot of their time on this. They do need to be aware of it. And trust me, your board is going to be interested in asking about these things.

The board question arrived on schedule. In-house counsel now tracks two things at once: what each state requires, and whether the requirement survives the federal pushback.

GC AI, the enterprise legal AI platform founded by a three-time general counsel (Anki, Bloomtech, and Replit) for in-house teams, treats that tracking load as a workflow. The Research feature returns cited answers from primary sources, the Files feature keeps the enacted texts and your memos in one permanent workspace, and Contract Intelligence, GC AI's newest launch, answers what your vendor contracts already promise about AI.

Here is the map.

The 2026 State AI Law Map at a Glance

Seven laws set the agenda, and two of them are in litigation. Every date below is verified against the enacted text or a primary announcement as of August 2026.

Law	Who It Covers	Effective	Status, August 2026
California SB 53 (Transparency in Frontier AI Act)	Developers training frontier models above 10^26 FLOPs; heaviest duties above $500M revenue	Jan 1, 2026	In effect
California AB 2013 (Training Data Transparency Act)	Developers of generative AI systems publicly available in California	Jan 1, 2026	In effect; court denied xAI's motion to block it, March 2026
California SB 942 (California AI Transparency Act)	Generative AI systems with 1M+ monthly users, publicly accessible in California	Aug 2, 2026	In effect
Texas TRAIGA	Mostly government deployers; narrow prohibited-use rules for the private sector	Jan 1, 2026	In effect
Illinois HB 3773	Employers using AI in employment decisions	Jan 1, 2026	In effect
Colorado AI Act (SB 26-189)	Developers and deployers of automated decision-making technology	Jan 1, 2027	Contested: enforcement suspended April 2026 amid xAI's suit and DOJ intervention
New York RAISE Act	Frontier model developers	Jan 1, 2027	Amended March 2026; on deck

California SB 53: The Transparency in Frontier AI Act

Governor Newsom signed SB 53, the Transparency in Frontier Artificial Intelligence Act (TFAIA), on September 29, 2025, and its core requirements took effect January 1, 2026. It is the first US law regulating frontier AI models, and it is deliberately narrow: it covers developers of foundation models trained above 10^26 FLOPs of compute, a threshold only the largest frontier labs cross, with the heaviest duties reserved for large frontier developers above $500 million in annual revenue. Covered developers must:

Publish a frontier AI framework describing how they assess and mitigate catastrophic risk
Publish transparency reports at deployment
Report critical safety incidents to California's Office of Emergency Services
Honor whistleblower protections for employees who flag catastrophic-risk concerns

Read the coverage threshold before you panic-file anything. SB 53 regulates the companies that train frontier models, a list measured in dozens, and your company is far more likely a customer of a covered developer than a covered developer.

The in-house payoff is indirect and useful: the law forced the safety paperwork into the open, and the first artifacts are already public. Anthropic published its compliance framework for the law in December 2025, and OpenAI released its Frontier Governance Framework in May 2026.

When your team evaluates a model vendor, those disclosures are public documents that belong in the vendor file next to the DPA.

California SB 942: The AI Transparency Act, In Effect Since August 2, 2026

SB 942, the California AI Transparency Act, is now the binding rule for a much wider set of companies. As amended by AB 853, it became operative on August 2, 2026 and covers providers of generative AI systems with more than a million monthly visitors or users that are publicly accessible in California.

Covered providers must:

Offer a free AI detection tool
Provide visible disclosure options for AI-generated content
Embed latent provenance disclosures, machine-readable markers identifying content as AI-generated, in the media itself

Penalties run $5,000 per violation, each day counts as a discrete violation, and city attorneys and county counsel can bring the action alongside the Attorney General.

Two reading notes for in-house teams. First, the million-user threshold means the direct obligations again land on your vendors, and your exposure runs through the license: covered providers must revoke the licenses of licensees who strip the disclosure tooling, which is how the obligation reaches companies that build products on top of covered models.

If August 2 came and went while the provenance work sat in a sprint backlog, treat it as this week's project, and read your license for the revocation trigger while you are in there.

Second, August 2 was also the day the EU AI Act's Article 50 transparency obligations began to apply, so a single content-labeling workstream can retire both.

The Rest of the Map: State AI Laws in Effect and on Deck

Five more laws fill in the 2026 map, and three of them moved this spring:

California AB 2013, in effect since January 1, 2026: The Training Data Transparency Act requires developers of generative AI systems to post public summaries of their training datasets across 12 categories. xAI sued to block it as a forced disclosure of trade secrets; a federal judge denied the preliminary injunction in March 2026, so the duty stands while the case continues. File those summaries with the SB 53 frameworks you already collected.
Texas TRAIGA, in effect since January 1, 2026: The Texas Responsible Artificial Intelligence Governance Act arrived far narrower than its first draft: prohibited-use rules on behavioral manipulation, social scoring, and certain biometric uses aimed mainly at government deployers, plus a sandbox program. Private-sector duties are thin, and intent standards keep enforcement narrow.
Illinois HB 3773, in effect since January 1, 2026: Illinois amended its Human Rights Act to make it a civil rights violation to use AI in employment decisions in a way that discriminates, with a notice requirement when AI is used in hiring. If HR runs AI screening on Illinois candidates, this one applies to you today.
Colorado AI Act, rewritten in May and suspended in April: On May 14, 2026, Governor Polis signed SB 26-189, which repealed the 2024 act, pushed the effective date to January 1, 2027, and swapped the duty of care, risk-management programs, and impact assessments for disclosure rules around automated decision-making. The litigation over the original act continues either way, and enforcement is paused; the preemption section below has the details. Any compliance memo written before May 2026 describes a law that no longer exists in that form.
New York RAISE Act, effective January 1, 2027: A chapter amendment on March 27, 2026 reshaped New York's frontier-model safety law, which now takes effect January 1, 2027 and gives New York its own version of the SB 53 conversation with different thresholds.

First drafts promise sweeping obligations, final texts narrow to disclosure and specific-harm rules, and effective dates slip. Track the enacted text, calendar the real dates, and re-verify before you brief anyone, because the tracker you read in January is already wrong twice over.

The Preemption Fight: Two Executive Orders, a DOJ Task Force, and a Live Case

On December 11, 2025, the White House issued an executive order on national AI policy aimed squarely at state AI laws. The order:

Directed the Department of Justice to build an AI Litigation Task Force for challenging state AI statutes in court
Told the Commerce Department to catalog state laws that conflict with federal policy
Conditioned certain federal funding on states holding back from onerous AI regulation

A second order followed on June 2, 2026, extending the federal framework push; Gibson Dunn's analysis covers the mechanics.

The task force stopped being an org chart in April. xAI had sued Colorado on April 9, 2026 in the US District Court for the District of Colorado, asking to enjoin the Colorado AI Act on First Amendment, Equal Protection, and Dormant Commerce Clause grounds.

On April 24, the DOJ filed a complaint in intervention, the federal government's first courtroom intervention against a state AI law, arguing the statute compels discrimination in violation of the Equal Protection Clause. Three days later the court granted a joint motion suspending enforcement, and Colorado's Attorney General has said enforcement will wait until his office finishes rulemaking.

Congress is moving on a slower clock. The Great American AI Act, a 269-page bipartisan discussion draft released June 4, 2026 by Representatives Jay Obernolte and Lori Trahan, would preempt state AI development laws for three years while leaving deployment rules and generally applicable state law untouched. The draft drew criticism from both parties, and the Senate Commerce Committee postponed its AI markup to September.

Preemption is the constitutional doctrine that valid federal law displaces conflicting state law, and it runs through Congress or valid federal regulation. An executive order sets the administration's litigation posture, and courts decide each state law's fate one case at a time.

The early results cut both ways: Colorado's act is suspended, while California's training-data law survived its first test when the court denied xAI a preliminary injunction. That sets up the compliance position that survives every outcome: comply with enacted state law that is in effect, and treat a preemption win as upside if it comes.

A company that skips SB 942 provenance work on a preemption bet is wagering product architecture on unresolved constitutional litigation.

For the board deck: five state laws bind today, one is suspended mid-lawsuit, two more arrive in January 2027, the federal government is fighting them in court and in Congress, and legal is complying with what is enforceable while tracking what is contested.

What In-House Counsel Should Do Now

Rachel Harris, General Counsel and AI Governance and Privacy Officer at Suzy and a CZ and Friends guest, has a name for compliance work that exists to look like compliance work:

If you don't do those exercises, you will end up in compliance theater, to where you are just putting words in a questionnaire for the sake of having the words in the questionnaire.

A seven-law year invites exactly that: the 50-state survey nobody reads, billed at outside-counsel rates.

GC AI CEO Cecilia Ziniti told the flip side of that story on the same episode: a GC AI customer that operates point-of-sale machines got a CCPA disclaimer from outside counsel that ran longer than the checkout screen, and the customer's lawyer worked directly with engineering to replace it with a consent flow of about five words, “tap to consent.”

The statute became a product decision. State AI law rewards the same posture, and the workload breaks into five concrete tasks:

Run one AI inventory for both regimes: The system inventory that anchors EU AI Act compliance answers the state-law questions too: what AI touches employment decisions (Illinois), what generates public-facing content (SB 942), and which vendors are covered developers whose SB 53 frameworks and AB 2013 training-data summaries belong in your diligence file.
Close the Illinois employment gap first: It is in effect, it reaches any Illinois hiring, and the fix, notice plus discrimination testing of AI screening tools, is a defined project with an end date.
Confirm SB 942 provenance survived your integration: The law is in force now. If any feature surfaces generative output from a covered provider, confirm the disclosure plumbing works in production this week, and read the license for the revocation trigger while you are in the document.
Ask your contracts what they promised about AI: The provenance question is a portfolio question: which vendor agreements license generative AI capability, and which carry disclosure or provenance obligations you have already accepted? Contract Intelligence loads those agreements into a Vault and extracts the AI-related provisions into a sortable View with a citation back to each source passage, which turns a week of tab-flipping into a filtered list.
Write the contested-law caveat once: Every AI-law memo you send should carry the same two sentences: the obligation is enforceable today, and xAI v. Colorado, the DOJ's intervention, and the preemption push in Congress may change that on a multi-year horizon. It keeps the board briefed without re-litigating strategy every news cycle.

Then put the re-verification on a schedule, and fold the results into the same corporate legal compliance program that owns your other regulatory calendars. Colorado changed shape in May, the federal posture has moved three times since December, and state AI law in 2026 has a shelf life measured in weeks.

GC AI's Regulatory Monitoring Skill Creator, a pre-built skill in the Skill Library, runs this as a standing automation: set the jurisdictions, the risk area, and a materiality threshold, and it scans legislative databases, verifies each flagged change against the bill text, and returns a prioritized action-items table on the schedule you set instead of the one you remember to run.

See a recurring monitor built and scheduled end to end:

How GC AI Helps You Track State AI Laws

The tracking duty belongs to the legal team, and no platform can discharge it for you. A legal AI platform changes the cost of staying current across 50 states and two continents. GC AI posted its widest margin on the In-House Legal Bench in regulatory tracking. The Bench is GC AI's own May 2026 benchmark of 100 real in-house tasks scored against 1,200+ attorney-developed criteria:

GC AI: 86.8%
ChatGPT (GPT-5.5): 79.8%
Claude (Opus 4.7): 68.4%
Gemini (3.1 Pro): 57.5%

The categories behind GC AI's largest advantages were regulatory tracking and legal research, which is exactly the work those seven laws demand. Four features do that work:

The Research feature deploys simultaneous agents against primary law and government sources and returns citations you can check
The US Case Law feature searches 13M+ US court opinions with treatment flags, so the litigation half of the tracker, xAI v. Colorado included, stays verifiable in the same workspace
The Files feature keeps the enacted texts and memos in one permanent workspace across quarters
The Regulatory Monitoring Skill Creator turns the quarterly sweep into a recurring agent with a priority tier on every finding

Watch the US Case Law feature research a forum-selection clause under Delaware law, from question to cited answer:

Wendra Liang, VP of Legal at Vercel, works at the speed these laws move:

Every day, our legal team depends on GC AI to enable us to move at the lightning speed of Vercel's business, and it's the first product I've felt is truly built for the kind of lawyer I aspire to be.

Speed like that keeps the board briefed in a year when one state law can be amended, delayed, and sued in the same quarter. The duties travel with the lawyer, too: ABA Formal Opinion 512 puts competence and confidentiality obligations on any lawyer who brings AI into the tracking workflow itself, the same AI legal ethics baseline that governs the rest of your practice.

In GC AI's December 2025 ROI study of more than 100 active customers, lawyers reported an average of 14 hours back per week. Start with one question you owe the board this quarter, and see how fast the cited answer comes back.

Watch Contract Intelligence take a portfolio question from upload to cited answer:

Start Asking: Try GC AI’s Contract Intelligence

Frequently Asked Questions

Is There a Federal AI Law in the United States?

There is no general federal AI statute as of August 2026, and the Great American AI Act remains an unpassed discussion draft. Federal policy runs through executive orders and agency action, so compliance means tracking each state's enacted law individually.

Which State AI Laws Are Currently in Effect?

Five state AI laws bind as of August 2026: California SB 53, California AB 2013, Texas TRAIGA, and Illinois HB 3773 since January 1, and California SB 942 since August 2. Colorado's act is suspended in litigation and New York's RAISE Act arrives January 1, 2027, so the count keeps moving; re-check it before you brief anyone.

Does SB 53 Apply to My Company?

Only if you train frontier models above the 10^26 FLOPs compute threshold, a list measured in dozens of companies. For everyone else the law works in your favor: covered developers now publish safety frameworks and transparency reports, which are free diligence material for your vendor file.

What Does SB 942 Require and Who Must Comply?

SB 942 reaches providers of generative AI systems with more than one million monthly users that are publicly accessible in California, and it reaches smaller companies indirectly through license terms, because covered providers must revoke the licenses of anyone who strips the disclosure tooling. If your product builds on a covered model, the compliance question is whether provenance survives your integration.

Can an Executive Order Preempt State AI Laws?

No, an executive order cannot itself preempt state law. Preemption requires an act of Congress or a valid federal regulation, and courts decide whether a specific state law falls. The December 2025 order instead directed the DOJ to challenge state AI laws in litigation, which it is now doing in xAI v. Colorado, so state obligations remain enforceable unless a court suspends them.

Should My Company Skip SB 942 Compliance While Preemption Litigation Is Pending?

No, SB 942 is in effect and enforceable now. Meet the enforceable obligations today and treat a preemption ruling in your favor as a future development; skipping the provenance work is a bet on unresolved constitutional litigation.

How Often Should In-House Teams Re-Verify Their State AI Law Tracker?

Quarterly at minimum, and after any major federal move: Colorado changed shape in May 2026 and the federal posture has shifted several times since December 2025. GC AI's Regulatory Monitoring Skill Creator runs the sweep as a standing automation that scans legislative databases and returns changes ranked by priority on the schedule you set.