Written and maintained by CASRAI Editorial Board
Published
On this page
AI regulation in the US is a layered system rather than one law: as described in the sources below on 2026-11-15, there is no comprehensive federal AI statute, the federal government acts mainly through executive orders and agency direction, and the binding AI-specific rules that exist are state laws such as Texas HB 149 (TRAIGA), California SB 53 and Colorado SB24-205. Which layer matters to you depends on whether you build, deploy or fund AI, and in which state.
What US AI regulation is and who enforces it
People searching for “us ai regulation” or “ai regulation laws” usually want one answer, and the honest answer has three parts. First, Congress has not passed a statute that regulates AI as a whole. Second, the executive branch sets federal AI policy through orders that direct agencies, and those orders do not themselves create duties for private developers. Third, individual states have enacted their own AI laws, each with its own scope, enforcer and penalties. “State AI regulation” is therefore where the enforceable obligations currently sit.
Enforcement differs by layer. A federal executive order is carried out by the agencies it names. A state statute is enforced by whoever that statute designates; in the two state laws read for this page, that is the state attorney general (Texas HB 149 gives the attorney general “exclusive authority to enforce this chapter”, and California’s governor described SB 53’s civil penalty as enforced by the Attorney General). There is no single national AI regulator.
Federal layer: executive orders, not a statute
EO 14179 (January 2025)
Executive Order 14179, “Removing Barriers to American Leadership in Artificial Intelligence”, was signed on January 23, 2025 and revoked the earlier Biden-era order EO 14110. As summarised in CASRAI’s guide to EO 14179 and EO 14365, it set a policy of sustaining US AI dominance and directed agency reviews, but it created no new requirement for companies; its practical effect was ending EO 14110’s reporting requirement for the largest dual-use foundation models.
EO 14365 (December 2025): the preemption push
Executive Order 14365, “Ensuring a National Policy Framework for Artificial Intelligence”, was signed on December 11, 2025. Its stated policy is a “minimally burdensome national policy framework for AI”, and its target is state law rather than AI companies. As read on the govinfo.gov text on 2026-11-15, it directs the following:
- Section 3: within 30 days the Attorney General must establish an AI Litigation Task Force to challenge state AI laws inconsistent with the order’s policy, on grounds that include interference with interstate commerce and federal preemption.
- Section 4: within 90 days the Secretary of Commerce must publish an evaluation identifying “onerous” state AI laws and laws to refer to that Task Force.
- Section 5: a Commerce (NTIA) Policy Notice making states with onerous AI laws ineligible for BEAD non-deployment funds “to the maximum extent allowed by Federal law”, and an agency assessment of whether discretionary grants can be conditioned on states not enacting or enforcing conflicting AI laws.
- Section 6: an FCC proceeding on a federal AI reporting and disclosure standard that would preempt conflicting state laws.
- Section 7: an FTC policy statement on how the FTC Act’s ban on unfair or deceptive practices applies to AI models, including when state laws that require alterations to models’ truthful outputs are preempted.
- Section 8: a legislative recommendation for a uniform federal AI framework that preempts conflicting state laws, which must not propose preempting otherwise lawful state laws on child safety protections, AI compute and data center infrastructure (other than generally applicable permitting reforms), or state government procurement and use of AI.
What an executive order can and cannot do
An executive order directs the executive branch; it is not a statute. This is an editorial reading, and it matches CASRAI’s guide, which states that an executive order cannot by itself preempt a state law. EO 14365 itself reflects this: its Section 8 asks for legislation, which would require Congress to act. Until a court, Congress or a state changes a state law, that law stays on the books, although funding conditions and litigation can add pressure on states. For the “ai regulation bill” question, the relevant item in the order is that legislative recommendation; the sources read for this page do not describe a specific federal bill.
State AI regulation: three examples with sourced detail
Texas HB 149 (TRAIGA)
The enrolled text of Texas HB 149, the Texas Responsible Artificial Intelligence Governance Act, takes effect on January 1, 2026 (Section 10). Read on 2026-11-15, it provides that:
- The attorney general has “exclusive authority to enforce this chapter” (Section 552.101); there is no private right of action.
- A 60-day notice-and-cure period applies before enforcement (Section 552.104).
- Civil penalties (Section 552.105) range from $10,000 to $12,000 per curable violation, $80,000 to $200,000 per uncurable violation, and $2,000 to $40,000 per day for continuing violations.
- State licensing agencies can sanction a licensee on the attorney general’s recommendation, with a monetary cap of $100,000 (Section 552.106).
- Chapter 553 creates a regulatory sandbox run through the Department of Information Resources, with participation of up to 36 months and quarterly reports.
CASRAI’s Texas TRAIGA explainer covers the statute’s duties in more depth.
California SB 53 (frontier developers)
The governor’s office release of September 29, 2025 announcing the signing of SB 53 describes a transparency law for large frontier developers: they must publish a framework on their website, a channel exists to report critical safety incidents to the California Office of Emergency Services, whistleblowers are protected, a civil penalty is enforced by the Attorney General, and a CalCompute consortium is created, with annual recommendations from the Department of Technology. Thresholds and deadlines are in the statute text, which this page does not restate; CASRAI publishes a California SB 53 guide.
Colorado SB24-205 (consumer protections)
The Colorado General Assembly’s bill page lists SB24-205, “Consumer Protections for Artificial Intelligence”, as signed on May 17, 2024, with a summary stating that duties apply “On and after February 1, 2026”. Because a bill page’s date can be superseded by later amendments, use CASRAI’s Colorado AI Act explainer for the current position rather than relying on the original date.
Three different kinds of state law
The examples show that “ai regulation laws” is not one category. Texas HB 149 is a general state AI governance statute with attorney-general enforcement and a sandbox. California SB 53 targets frontier developers with publication, incident-reporting and whistleblower provisions. Colorado SB24-205 is framed as consumer protection. For a side-by-side of the three states’ approaches, see California vs Colorado vs New York AI law; Utah and New York have their own explainers (Utah, New York RAISE Act).
Compliance steps: working out which layer applies
These steps are an editorial decision aid, not legal advice, and each depends on reading the primary text.
- Define your role. Building or releasing a model, deploying a tool that affects people, or receiving federal or state money are different positions, and the laws read above attach to different ones (for example SB 53 to large frontier developers).
- List the states where the system operates or its outputs reach people. State statutes are enforced by their own attorneys general, so state-by-state is the unit of analysis.
- Read the enrolled statute, not a summary. Definitions, effective dates and penalties are in the text; the Texas figures above come from the enrolled bill.
- Track federal preemption efforts separately. EO 14365 creates a litigation task force, a Commerce evaluation and funding conditions, but none of these is a court ruling that a given state law is invalid.
- Adopt a risk-management framework as internal evidence. The NIST AI RMF is a voluntary framework that can serve as a reference point for organising AI governance.
- Record decisions and dates. Because the landscape changes, note which source and date each compliance conclusion rests on.
Typical timeline phases
The sources describe a sequence rather than a single calendar. For the federal preemption effort, EO 14365 sets relative deadlines: a task force within 30 days of the order, then a Commerce evaluation, a funding Policy Notice and FCC and FTC actions on 90-day clocks, with legislation to follow if Congress chooses to act. For state laws, the phases are enactment, an effective date (Texas HB 149: January 1, 2026), a cure period where the statute provides one (Texas: 60 days), and enforcement by the attorney general. Texas also provides a time-limited sandbox of up to 36 months for approved testing.
Common mistakes
- Treating an executive order as a law that overrides states. EO 14365 directs agencies and asks for legislation; it does not itself void any state statute.
- Assuming there is one national rulebook. The two state laws read for this page differ in who they cover, who enforces them and how penalties work.
- Quoting a penalty without its category. Texas penalties depend on whether a violation is curable, uncurable or continuing.
- Relying on a signing date as an effective date. Colorado’s bill was signed on May 17, 2024; the duties on the bill page apply from a later date, and later changes should be checked in the current explainer.
- Ignoring the funding angle. EO 14365 ties BEAD non-deployment funds and possibly discretionary grants to state AI laws, which matters to institutions that rely on state or federal money.
Related instruments at a glance
The data table below compares the instruments read for this page by level, type and key mechanism. For the global picture, including how US rules compare with other regimes, see AI regulations around the world.
Frequently asked questions
Is there a federal AI law in the US?
No comprehensive federal AI statute is described in the sources read for this page as of 2026-11-15. Federal policy is set mainly through executive orders such as EO 14179 and EO 14365, which direct agencies rather than regulate developers directly.
Does EO 14365 override state AI laws?
Not by itself. It creates an AI Litigation Task Force, a Commerce evaluation of state laws, funding conditions and a legislative recommendation for preemption. An executive order is not a statute, and state laws remain in force unless a court, Congress or the state changes them.
What does state AI regulation look like in practice?
It varies by state. Texas HB 149 is enforced exclusively by the attorney general with a 60-day cure period and set penalty ranges; California SB 53 imposes publication and incident-reporting duties on large frontier developers; Colorado SB24-205 is a consumer-protection statute.
Which AI regulation bill is moving in Congress?
The source read for this page, EO 14365 Section 8, calls for a legislative recommendation for a uniform federal framework that preempts conflicting state AI laws, excluding areas such as child safety, AI compute and data center infrastructure, and state government procurement and use of AI. It does not name a specific bill.
Who enforces these laws?
For the state statutes read here, the attorney general: Texas gives the attorney general exclusive enforcement authority, and California’s SB 53 civil penalty is enforced by the Attorney General.
Sources
- Executive Order 14365, “Ensuring a National Policy Framework for Artificial Intelligence”, govinfo.gov, https://www.govinfo.gov/content/pkg/DCPD-202501186/html/DCPD-202501186.htm, read 2026-11-15.
- Texas HB 149 (89th Legislature), enrolled text, https://capitol.texas.gov/tlodocs/89R/billtext/html/HB00149F.htm, read 2026-11-15.
- Office of Governor Gavin Newsom, “Governor Newsom signs SB 53”, September 29, 2025, https://www.gov.ca.gov/2025/09/29/governor-newsom-signs-sb-53-advancing-californias-world-leading-artificial-intelligence-industry/, read 2026-11-15.
- Colorado General Assembly, SB24-205 “Consumer Protections for Artificial Intelligence”, https://leg.colorado.gov/bills/sb24-205, read 2026-11-15.
- CASRAI, “Trump AI executive orders: EO 14179 and EO 14365”, /guides/trump-ai-executive-orders-eo-14179-eo-14365, read 2026-11-15.
Data: US AI governance instruments compared by level, type and key mechanism
Compares five US federal and state AI instruments by level, legal type, signing or effective date and key mechanism, compiled from the primary texts and official releases listed in the sources, read 2026-11-15.
Period covered: 2024/2026. Coverage: United States.
| Instrument | Level | Legal type | Date stated in source | Key mechanism or enforcer |
|---|---|---|---|---|
| EO 14179 | Federal | Executive order | Signed 2025-01-23 | Revokes EO 14110; directs agency reviews; no new company requirement (per CASRAI guide) |
| EO 14365 | Federal | Executive order | Signed 2025-12-11 | AI Litigation Task Force, Commerce evaluation of state laws, BEAD funding condition, FCC and FTC actions, preemption legislative recommendation |
| Texas HB 149 (TRAIGA) | State (Texas) | Statute | Effective 2026-01-01 | Attorney general exclusive enforcement; 60-day cure; civil penalties $10,000 to $200,000 per violation plus daily amounts; sandbox up to 36 months |
| California SB 53 | State (California) | Statute | Signed 2025-09-29 | Frontier developer framework publication, critical safety incident reporting to Cal OES, whistleblower protection, civil penalty enforced by the Attorney General |
| Colorado SB24-205 | State (Colorado) | Statute | Signed 2024-05-17 | Consumer protections for AI; bill page summary states duties apply on and after 2026-02-01 |
Sources
- Executive Order 14365, govinfo.gov (accessed 2026-11-15)
- Texas HB 149 enrolled text (accessed 2026-11-15)
- Governor Newsom signs SB 53 (accessed 2026-11-15)
- Colorado SB24-205 bill page (accessed 2026-11-15)
- CASRAI guide to EO 14179 and EO 14365 (accessed 2026-11-15)
Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.







