Prolepsia forecastYes / no
Will the US Congress pass a federal law that limits state AI laws before 31 December 2027?
As of 15 November 2026, Prolepsia puts the chance of yes at 59%.
59%chance of yes
How Prolepsia reasons
The report Prolepsia wrote with this forecast.
How forecasts are made · Track record
TL;DR
The forecast is 59.2% Yes and 40.8% No for enactment of a qualifying federal statute between November 15, 2026, and December 30, 2027. The Yes advantage rests principally on narrow digital-replica and chatbot legislation, not on Congress adopting a general ban on state AI regulation.
Key evidence
- Broad preemption has a poor recent precedent. The Senate voted 99–1 on July 1, 2025, to remove the proposed state-AI moratorium. That is powerful evidence against a sweeping prohibition, but not a vote against every limited preemption clause (official roll call).
- NO FAKES supplies a concrete narrow route. The June 24, 2026 Senate-reported text expressly preempts specified state digital-replica claims, while preserving older laws and certain exceptions. This appears to fit the question’s application-specific scope (reported bill). However, Cruz blocked a September 30 passage attempt over speech and parody protections (Congressional Record).
- Child-safety legislation offers another route. The KIDS Act passed the House 267–117 on June 29, 2026 (House vote). Its text combines AI-chatbot obligations with express conflict preemption and exceptions for stronger state protections. Whether its package-wide wording meets the AI-specific requirement remains less clear (House-passed text).
- The political environment cuts both ways. Democrats won both incoming congressional chambers (AP), constraining sweeping preemption. The administration continues to seek statutory limits on state AI laws (White House recommendations).
Uncertainties
The decisive unknown is whether any final law retains an explicit, binding restriction on state AI authority. Federal safeguards alone do not qualify. Nor do executive action, committee approval, or passage in only one chamber.
NO FAKES still faces a substantive speech dispute. Chatbot proposals can change during negotiations, and the adopted CHATBOT substitute’s preemption language remains unverified. A compromise preserving all state authority would resolve No even if Congress passes substantial AI legislation. Enactment—not passage or implementation—must occur by December 30, 2027, at 11:59 p.m. Eastern Time.
Read the full report
TL;DR
The forecast is 59.2% Yes and 40.8% No for a qualifying federal law enacted between November 15, 2026, and December 30, 2027. The central route is a narrow restriction on state digital-replica or chatbot rules, not a nationwide ban on state AI regulation. Watch the final statutory preemption language and enactment date, rather than bill titles, executive announcements, or one-chamber votes.
Context
The question asks whether Congress and the president complete a new law that explicitly limits state authority over AI. A restriction covering one application or one regulatory requirement is enough. Federal standards without express state-law restrictions are not enough. The deadline is December 30, 2027, at 11:59 p.m. Eastern Time; a later effective date does not matter if enactment occurs within the window.
The political conflict is between national uniformity and states’ ability to impose their own protections. The administration’s legislative recommendations seek preemption of burdensome state AI laws (White House recommendations). Democrats won both incoming congressional chambers, changing the coalition needed for most of the forecast period (AP election account). This makes a negotiated safety law with limited preemption a more credible route than sweeping deregulation.
Evidence
The historical backbone counsels against confusing legislative momentum with enactment. The 2022 American Data Privacy and Protection Act failed to become law despite a 53–2 committee vote (legislative record). The 2024 American Privacy Rights Act remained at referral. The 2024 Kids Online Safety and Privacy package failed to become law despite 91–3 Senate approval. These are selected cautionary comparisons, not a representative enactment base rate. Their lesson is that bipartisan support and success in one chamber do not settle disputes between chambers.
There is also a positive analogue. TAKE IT DOWN became law on May 19, 2025, demonstrating that narrow legislation addressing deepfake harms can reach enactment (enacted law). It cannot resolve this question because it predates the forecast window. It does show why application-specific legislation deserves separate attention from comprehensive AI policy.
The clearest direct precedent for broad preemption is negative. On July 1, 2025, the Senate voted 99–1 to remove the proposed AI moratorium from reconciliation legislation (official roll call). Opposition crossed party lines. I read this as strong evidence against reviving an expansive moratorium without an acceptable replacement regime. It is not evidence that senators reject every narrow restriction on state law.
The strongest narrow route is NO FAKES. Its June 24, 2026 Senate-reported text contains express preemption of specified state causes of action involving digital replicas of voices or visual likenesses. It preserves qualifying laws existing on January 2, 2025, and exceptions for certain election-related and sexually explicit replicas (reported text). Those carve-outs reduce the provision’s reach. They do not remove its restriction on other state claims.
My interpretation is that this restriction fits the question’s coverage of particular AI applications. There is still an interpretive issue because the bill’s digital-replica definition also reaches some non-AI technology. That uncertainty warrants caution, not exclusion of the route.
NO FAKES has moved beyond introduction, but it has not cleared its central obstacle. On September 30, 2026, Blackburn sought unanimous-consent passage and Cruz objected over speech and political-parody protections. Cruz also said he wanted a revised bill to pass (Congressional Record). This is evidence of an active negotiation with a concrete drafting problem. It is neither Senate passage nor proof that the disagreement is permanent.
Child-safety legislation offers another developed route. The House passed the KIDS Act 267–117 on June 29, 2026, including 104 Democratic votes (House Clerk). Its House-passed text places AI-chatbot obligations alongside an express conflict-preemption clause, while preserving stronger state protections for minors and specified state-law claims (House-passed text). The package vote establishes political compatibility with that combination, not separate approval of every preemption provision.
The KIDS route has a wording problem as well as a legislative one. Its preemption provision is package-wide rather than a clause directly naming state AI requirements. An express clause applying to an AI-specific title has a credible claim to qualification, but the final text must make that connection sufficiently explicit. Stronger-protection exceptions do not automatically prevent qualification because the question allows conflict preemption.
Senate Commerce also advanced the CHATBOT Act on August 5, 2026 (committee announcement). Its introduced text contains express conflict preemption with protection for stronger state rules (introduced bill). The relevant language in the adopted substitute remains unverified. Committee advancement therefore supports legislative interest, but does not establish that the current proposal retains a qualifying clause.
A frontier-model bargain is another route, though less advanced. The July 23, 2026 FRONTIER Act expressly restricts new state developer obligations concerning catastrophic-risk transparency, independent evaluation, and incident reporting. Its introduced text is narrower than a blanket moratorium and does not retain the earlier discussion draft’s three-year sunset (introduced bill). The recorded progress remains committee referral rather than chamber passage (action record).
Later reporting weakens reliance on that original bargain. November 10 reporting says Trahan is reworking the proposal without Obernolte toward minimum federal standards that states can exceed (Spectrum News NECN). A federal floor can still contain qualifying conflict preemption, but the replacement language is not established. The older draft is evidence of a negotiating approach, not an agreed destination.
What's non-obvious
A law that preserves most state authority can still resolve Yes. The question does not require Congress to extinguish state AI regulation. A safety statute can preserve tort claims, stronger protections, and whole categories of state law while explicitly displacing a smaller set of AI-specific requirements. The digital-replica and chatbot texts show how this can happen. Reading every state-law exception as a defeat for preemption misses the restriction that remains.
The opposite mistake is to treat every federal AI bill as progress toward Yes. Some proposals preserve state authority without imposing an affirmative limit. Others regulate activities involving AI but use only general preemption language. The decisive information is in the operative clause, not the policy label. Several pending bills also share the same floor-time constraints, negotiators, and cross-chamber disputes. Multiple vehicles broaden the available routes; they do not eliminate the common bottlenecks.
Uncertainties
The largest uncertainty is whether qualifying language survives compromise. NO FAKES must resolve speech objections without losing the relevant state-law restriction. Child-safety negotiations must reconcile substantive protections and preemption. A frontier-model agreement must satisfy lawmakers seeking enforceable national safeguards and an administration seeking limits on conflicting state rules.
The adopted CHATBOT substitute is a specific evidence gap. Its authoritative current language would establish whether that vehicle still offers express preemption or only preserves state law. Official replacement text for the revised frontier proposal would likewise clarify whether a qualifying restriction remains.
There is no verified agreed qualifying defense or appropriations rider in the evidence reviewed. That is not proof that negotiations or amendments contain none. Larger packages remain possible vehicles, but their importance does not guarantee that a contested AI provision survives.
Finally, enactment timing matters independently of policy agreement. Pending proposals need successor vehicles if they do not pass before the congressional reset. Both chambers must approve the same qualifying language, and the bill must become law within the window. These obstacles explain the substantial No probability while leaving a modest Yes advantage driven by narrow legislation.
Forecast history
| Date | Forecast |
|---|---|
| 59% chance of yes |
Sources Prolepsia read
- legislative recordapi.congress.gov
- U.S. Senate: U.S. Senate Roll Call Votes 119th Congress - 1st Sessionwww.senate.gov
- reported textwww.govinfo.gov
- Office of the Clerk, U.S. House of Representativesclerk.house.gov
- Cruz, Schatz’s CHATBOT Act Advances to the Senate Floor - U.S. Senate Committee on Commerce, Science, & Transportationwww.commerce.senate.gov
The question
This question asks whether the United States will enact a new federal statute that explicitly restricts the authority of U.S. states to enact, implement, or enforce laws regulating artificial intelligence (AI) between November 15, 2026, and December 30, 2027, inclusive. The restriction may take the form of federal preemption of state AI laws, a temporary moratorium on state AI regulation, or another binding statutory limitation on state legislative or regulatory authority over AI. As of November 15, 2026, federal preemption of state AI laws remains a major unresolved policy issue. In July 2025, the U.S. Senate voted 99–1 to remove a proposed ten-year moratorium on state AI regulation from a federal budget reconciliation bill. The Trump administration subsequently pursued a national AI regulatory framework through executive action and legislative recommendations. In June 2026, a bipartisan congressional proposal outlined a three-year preemption of certain state laws governing AI model development. However, no general federal statute preempting state AI regulation had been enacted as of late October 2026. Sources include [Roll Call](https://rollcall.com/2026/06/04/bipartisan-ai-draft-proposes-three-year-preemption-of-state-laws/) and [the federal judiciary's 2026 AI policy discussion materials](https://ord.uscourts.gov/2026%20District%20Conference/Effective%20AI%20Panel%20-%20Your%20AI%20is%20on%20Display.pdf). The question concerns whether Congress successfully establishes a new statutory restriction on state AI regulatory authority before the specified deadline, regardless of whether the legislation establishes a comprehensive federal AI regulatory framework. It includes legislation passed as a standalone bill or as part of a broader legislative package, such as an appropriations bill or a national defense authorization act. The forecast begins on November 15, 2026, and the deadline is December 30, 2027, at 11:59 p.m. Eastern Time. The question will be evaluated based on federal legislation enacted by that deadline.
How it resolves
The question resolves YES if, between November 15, 2026, and December 30, 2027, inclusive, a bill passes both chambers of the U.S. Congress and becomes federal law through presidential signature, congressional override of a presidential veto, or another constitutionally valid enactment mechanism, and the enacted statute contains at least one explicit, legally binding provision that restricts the authority of U.S. states to enact, implement, or enforce laws or regulations specifically governing artificial intelligence. Qualifying provisions include an express federal preemption clause invalidating or superseding specified categories of state AI laws; a temporary or permanent moratorium prohibiting states from enacting or enforcing specified AI regulations; or an explicit statutory prohibition on states imposing certain AI-specific regulatory requirements, such as licensing requirements for AI developers, mandatory AI safety assessments, or AI model transparency obligations. A provision qualifies even if its scope is limited to a particular category of AI systems, AI applications, or AI regulatory requirements. It does not need to preempt all state AI laws or eliminate state authority over AI generally. A law that expressly preempts only state requirements conflicting with a newly established federal AI regulatory standard also qualifies, provided that the statute explicitly identifies AI-related state regulation as subject to that restriction. The question resolves NO if no qualifying federal statute is enacted by the deadline. This includes circumstances in which legislation passes only one chamber, passes both chambers but is not enacted, or is introduced or approved in committee without becoming law. It also resolves NO if the federal government attempts to restrict state AI laws exclusively through executive orders, agency regulations, administrative guidance, litigation, or judicial decisions without enactment of a qualifying federal statute during the forecast period. A general federal statute concerning AI that does not explicitly restrict state regulatory authority does not qualify. Neither does a statute containing only a general savings clause preserving state authority, or a statute that establishes federal AI standards without explicitly preempting or prohibiting state AI requirements. Incidental preemption arising solely from a court's interpretation of otherwise silent federal legislation is insufficient. The primary resolution sources will be the official enacted statutory text published by Congress.gov and the U.S. Government Publishing Office, supplemented by official congressional records and White House enactment announcements. The controlling evidence will be the text of the enacted law rather than descriptions in press releases or media reporting. If a qualifying statute is enacted before the deadline but its relevant provisions have a delayed effective date, the question resolves YES because enactment, rather than implementation, is the relevant event. If the statute is subsequently challenged in court or temporarily enjoined, it still qualifies, provided it was validly enacted by the deadline. Legislation enacted after December 30, 2027, does not qualify, even if Congress passed it before that date. If official publication of the enacted statutory text is delayed, resolution may be postponed until authoritative documentation becomes available. The question should not be annulled merely because Congress does not enact qualifying legislation; that outcome resolves NO.
Fine print
For this question, 'state AI laws' means statutes, regulations, or legally binding requirements adopted by any of the 50 U.S. states or the District of Columbia that specifically regulate artificial intelligence systems, their development, deployment, operation, or use. Restrictions on local governments also qualify if the statute explicitly restricts state-level AI regulatory authority, but restrictions applying exclusively to municipal or county governments do not qualify. A general federal law that incidentally affects AI-related activities, such as an ordinary tax law, telecommunications law, or general consumer protection law, does not qualify unless it expressly restricts state authority over AI regulation. Federal laws that merely create financial incentives for states to adopt particular AI policies do not qualify unless they establish a binding legal restriction on state AI regulatory authority. Similarly, nonbinding congressional resolutions, policy statements, recommendations, and voluntary federal-state agreements do not qualify. A qualifying restriction can be narrow, temporary, or subject to exceptions. For example, legislation prohibiting states from imposing licensing requirements on developers of frontier AI models would qualify even if states retained authority over other AI applications. However, the relevant restriction must be explicit in the enacted statutory language. The enactment date, rather than the date of congressional passage or the effective date of the relevant provision, determines whether the deadline is met. The question covers only statutes newly enacted during the forecast period, not previously enacted laws or subsequent reinterpretations of existing legislation. The deadline is interpreted strictly as before December 31, 2027, meaning enactment must occur no later than December 30, 2027, at 11:59 p.m. Eastern Time.
More public forecasts
Will a US federal law that overrides state AI laws be in force on 31 December 2027?
21%chance of yes
Prolepsia forecast
Will Congress pass a law limiting the president's power to impose tariffs before 31 December 2028?
32%chance of yes
Prolepsia forecast
Will the US Congress pass a crypto market structure law, such as the CLARITY Act, by 31 December 2027?
29%chance of yes
Prolepsia forecast
Will the US Congress pass changes to clean energy tax credits before 31 December 2027?
50%chance of yes
Prolepsia forecast
Will Congress pass all 12 regular appropriations bills for fiscal year 2028 before 1 October 2027?
2%chance of yes
Prolepsia forecast
Will any US state enact a law imposing a statewide moratorium on new data centres between 1 January and 31 December 2027?
98%chance of yes
Prolepsia forecast