Prolepsia forecastYes / no
Will a US federal law that overrides state AI laws be in force on 31 December 2027?
As of 15 November 2026, Prolepsia puts the chance of yes at 21%.
21%chance of yes
How Prolepsia reasons
The report Prolepsia wrote with this forecast.
How forecasts are made · Track record
TL;DR
The forecast is 20.8% Yes and 79.2% No for a qualifying federal statute to be legally in force on December 31, 2027. The most plausible Yes outcome is targeted preemption within a federal frontier-AI safety law, not a blanket ban on state AI regulation.
Key evidence
- The strongest legislative test failed decisively. On July 1, 2025, the Senate voted 99–1 to remove the proposed state-AI moratorium. That demonstrates resistance to broad preemption, though not necessarily to narrower compromises.
- Concrete qualifying language exists. Section 9 of the introduced FRONTIER Act restricts state developer obligations concerning catastrophic-risk transparency, independent auditing or verification, and incident reporting. Its preservation of other state powers does not erase that substantial categorical preemption. The bill has 11 cosponsors, but its recorded actions show introduction and referrals rather than committee passage; the record’s update vintage is October 2, 2026.
- The political direction has become less favorable. Democratic control of both incoming congressional chambers makes cross-party agreement essential. November 10 reporting says Democratic FRONTIER negotiator Lori Trahan is moving toward federal minimum standards that states could exceed.
- Presidential demand remains clear. The White House’s March 20, 2026 framework supports preemption. It remains a recommendation, not an enacted restriction on states.
Uncertainties
The central unknown is whether stronger federal safeguards can secure Democratic agreement to limited regulatory exclusivity. Anthropic’s framework supports that exchange conditionally, but corporate support is not a congressional agreement.
Trahan’s replacement text is not available, so a move toward a federal floor does not establish that every qualifying preemption clause will disappear. A negotiated bill or rider remains a real risk to the No forecast. Even then, enactment alone is insufficient: the provision must cover a substantial AI-specific category, take effect by the deadline, and remain legally operative. Executive action cannot independently satisfy the question.
Read the full report
TL;DR
The forecast is 20.8% Yes and 79.2% No for a qualifying federal override of state AI laws to be legally in force on December 31, 2027. The central obstacle is agreement between Congress and the President on which state protections federal law would replace. Watch negotiated statutory text and its effective date, not executive announcements or general support for national AI rules.
Context
As of November 15, 2026, Congress has not enacted the general AI-preemption framework at issue. The administration supports national uniformity: Executive Order 14365, signed December 11, 2025, directs executive action and legislative recommendations, while the March 20, 2026 White House framework recommends preempting certain state AI requirements. Neither document supplies the enacted statute this question requires.
The threshold is narrower than a comprehensive federal takeover of AI regulation. A law reserving a substantial category of frontier-model auditing or reporting requirements to the federal government would qualify even if most state authority survived. But the restriction must be express, enacted, effective by the deadline, and legally operative at the end of December 31, 2027. A federal minimum that leaves states free to add requirements does not, by itself, meet that test.
Evidence
The historical backbone is the Senate’s July 1, 2025 vote of 99–1 to remove the proposed AI moratorium. This is the closest direct legislative test in the evidence. It shows that strong administration interest did not translate into support for a broad restriction on state AI regulation. It is one failed proposal, however, not a measured failure rate for all forms of preemption. A negotiated replacement regime presents lawmakers with a different choice from a moratorium.
Other technology-policy precedents point in both directions. The American Data Privacy and Protection Act received a 53–2 committee vote in 2022 but did not become law. The SELF DRIVE Act passed the House in 2017 but also failed to become law. Conversely, the Lautenberg Act became law on June 22, 2016, combining stronger federal regulation with scoped state preemption. These selected examples establish that such bargains are possible and that bipartisan progress can still stall. They do not provide a defensible statistical base rate for this question.
The strongest affirmative evidence is actual qualifying language. The FRONTIER Act, H.R. 9925, introduced July 23, 2026, contains express preemption in Section 9. It covers state developer obligations concerning catastrophic-risk transparency, independent auditing or verification, and incident reporting. The section preserves other state powers, but those exceptions do not negate the substantial category it reserves to federal authority. In this analysis, that provision would meet the substantive resolution threshold if enacted and operative.
This matters because the Yes case does not depend on inventing a legislative design. A limited safety-for-uniformity exchange has already been drafted. Nor does the introduced Section 9 expressly postpone preemption until the replacement federal regime becomes operational. Deadlines for federal rulemaking or company compliance should not automatically be treated as the preemption clause’s effective date.
The bill’s political standing is weaker than its substantive fit. Official records show 11 cosponsors, comprising six Democrats and five Republicans, alongside Republican sponsor Jay Obernolte. The recorded action history contains introduction and committee referrals, not committee passage. The record’s October 2, 2026 update vintage limits what it establishes about subsequent activity. This is evidence of a small bipartisan coalition, not an assembled passage coalition.
The election adds another constraint. AP’s election-call chronology establishes incoming Democratic control of both chambers. The main legislative window therefore requires agreement between Democratic congressional majorities and President Trump. Interest in federal AI safeguards does not resolve the dispute over state authority. Lawmakers can favor national standards while opposing a ceiling on stronger state protections.
The clearest recent signal concerns a negotiator’s behavior. Spectrum News reported on November 10 that Lori Trahan was preparing a replacement for FRONTIER without Obernolte, moving toward federal rules that states could exceed. Her office connected that change to the election. Replacement statutory text remains unavailable, so the report does not establish the fate of every preemption clause. It does mean the original bipartisan sponsor list should not be read as current agreement on Section 9.
There is still a bargaining zone. Anthropic’s June 2026 framework accepts narrowly specified preemption of frontier-governance requirements if Congress supplies sufficiently strong federal protections. Cantwell’s October 7 framework calls for enforceable standards, independent testing, and continuing oversight. Her statement does not commit her to preemption. Together, these positions identify the potential exchange, not an agreement: federal safeguards in return for limited regulatory uniformity.
Alternative legislative vehicles preserve a Yes pathway, but their details matter. The lame-duck legislative outlook identifies funding, defense, and possible children’s-safety or AI packages as opportunities. A moving vehicle is not evidence of an agreed preemption rider. The separate American AI Security Act, H.R. 10981, is another proposal to watch, but its substantive preemption terms are not established strongly enough here to carry additional weight.
Effective dates also eliminate apparent shortcuts. The introduced SECURE Data Act delays its general provisions, including preemption, until two years after enactment. That unchanged timing would miss this question’s deadline for a future enactment. The KIDS Act’s Senate-referred text preserves stronger protections for minors and generally delays effectiveness for one year. These examples show why passage of an AI-related law is not enough to establish Yes.
What's non-obvious
The decisive distinction is not federal action versus inaction. It is a federal floor versus a federal ceiling. Congress can enact substantial AI safeguards while leaving this question at No. Conversely, it can leave most state AI authority untouched and still produce Yes by expressly excluding states from a meaningful class of developer testing, auditing, or reporting requirements. The broad moratorium’s defeat therefore weighs against Yes without closing its most plausible route.
Preemption can also serve as the price of regulation rather than its opposite. Some developers accept limited national exclusivity alongside strong federal safeguards. That creates room for a bargain that differs from blanket deregulation. The inference is that the key signal is jointly accepted language defining exclusive federal functions—not another endorsement of a “national framework.” Current evidence shows the possible exchange more clearly than it shows the coalition needed to enact it.
Uncertainties
The largest unresolved issue is whether Democratic negotiators will accept any substantial ceiling on state requirements, and whether the President will accept the safeguards demanded in return. Trahan’s replacement text and an agreed Senate proposal would clarify that trade. Their absence is a gap in the evidence, not proof that negotiations cannot succeed.
The legislative record is also incomplete as a guide to current bargaining. The FRONTIER metadata have an October vintage, and the evidence does not establish a leadership-backed agreement or a qualifying-preemption vote count. Published replacement text, committee approval, and explicit White House acceptance would be more informative than additional hearings or general statements of support.
Finally, legal timing is a separate hurdle. Any qualifying provision must commence by the deadline and avoid expiration, repeal, invalidation, or a nationwide suspension that leaves it inoperative at resolution. An AI incident could accelerate a federal bargain, but it could also strengthen demands to preserve state powers. These uncertainties sustain a meaningful Yes possibility while leaving No as the clear forecast.
Forecast history
| Date | Forecast |
|---|---|
| 21% chance of yes |
Sources Prolepsia read
- White House National Policy Framework for AI - Legislative Recommendationswww.whitehouse.gov
- FRONTIER Act, H.R. 9925, introduced July 23, 2026www.govinfo.gov
- 11 cosponsorsapi.congress.gov
- Calling control of Congress in the 2026 midterm elections | The Associated Presswww.ap.org
- Rep. Trahan is replacing bipartisan AI safety billspectrumnewsnecn.com
- **Catastrophic Risk.** Consistent with existing US state law, “Catastrophic Risk” means a foreseeable and material risk that a Covered Developer’s development, storage, use, or deployment of a covered model will materially contribute to significant death, injury, or damage.²www-cdn.anthropic.com
- American AI Security Act, H.R. 10981www.congress.gov
- To direct the Secretary of Commerce to promulgate regulations related … | govscent.orggovscent.org
The question
This question asks whether, on December 31, 2027, at least one enacted United States federal statute is legally in force that expressly preempts, overrides, or prohibits the enforcement of a substantial category of state laws specifically regulating artificial intelligence (AI). The qualifying federal statute must establish a binding nationwide restriction on states' authority to regulate AI development, deployment, or use. It need not establish a comprehensive federal AI regulatory framework or prohibit every type of state AI regulation. As of November 15, 2026, Congress has not enacted a general federal statute preempting state AI regulation. In July 2025, the Senate voted 99–1 to remove a proposed ten-year moratorium on state AI regulation from federal budget legislation. On December 11, 2025, President Donald Trump signed Executive Order 14365, directing federal agencies to challenge certain state AI laws and develop recommendations for federal legislation establishing a uniform national AI policy. In March 2026, the White House proposed a legislative framework that would preempt certain state AI requirements. However, the executive order does not itself establish statutory preemption, and Congress has not enacted the proposed general preemption framework. Sources include [Executive Order 14365](https://www.whitehouse.gov/wp-content/uploads/2025/12/eo-14365.pdf) and [the Senate Commerce Committee's analysis of the proposed AI moratorium](https://www.commerce.senate.gov/wp-content/uploads/media/doc/AI%20Moratorium%20and%20Impact%20on%20State%20Laws%20%28002%29.pdf). The forecast covers federal legislative developments between November 15, 2026, and December 31, 2027. Its purpose is to assess whether Congress will establish legally binding federal supremacy over a meaningful portion of state AI-specific regulation by the end of 2027, rather than merely proposing legislation or pursuing preemption through executive action.
How it resolves
The question resolves YES if, on December 31, 2027, at least one United States federal statute enacted by Congress is legally in force and contains an express provision that prohibits states from enacting or enforcing, or otherwise supersedes, a substantial category of state laws or regulations specifically governing AI development, deployment, or use. To qualify, the federal statute must satisfy all of the following conditions: It must have completed the federal legislative process and become law through presidential signature, congressional override of a veto, or another constitutionally valid enactment procedure. Its relevant preemption provision must have taken legal effect on or before December 31, 2027. The provision must expressly restrict state legislative or regulatory authority over AI as a technology or over a meaningful class of AI systems, developers, deployers, or applications. The restriction must apply across multiple states rather than exclusively to a single state or locality. Qualifying legislation could include a temporary federal moratorium on state AI regulation, a federal AI regulatory framework expressly superseding conflicting state AI requirements, or a statute explicitly prohibiting states from imposing specified categories of AI-specific obligations on developers or deployers. The restriction need not cover all state AI laws. For example, a federal statute expressly preventing states from imposing independent safety-testing, risk-assessment, or reporting requirements on developers of general-purpose or frontier AI models would qualify, even if states retained authority to regulate AI in other contexts. The question resolves NO if no qualifying federal statutory provision is legally in force on December 31, 2027. This includes situations where Congress has introduced, debated, or passed legislation that has not become law; where legislation has been enacted but its relevant preemption provisions have not yet taken effect; or where a previously effective preemption provision has expired or been repealed before the resolution date. An executive order, federal agency regulation, administrative interpretation, or judicial decision does not independently qualify unless the qualifying preemption is expressly established by an enacted federal statute. Existing federal statutes that incidentally preempt particular state requirements through general conflict-preemption principles do not qualify unless they contain an express AI-specific preemption provision. A federal statute addressing only an isolated application of AI, such as a narrow rule concerning AI-generated election advertisements, does not qualify unless its preemption provision also applies to a substantial category of AI regulation extending beyond that isolated application. Resolution will be based primarily on the official text of enacted federal laws published through Congress.gov, the United States Statutes at Large, and the Office of the Law Revision Counsel's United States Code. Effective-date provisions, subsequent amendments, repeals, and relevant binding federal court decisions will also be considered. If sources disagree, the enacted statutory text and legally controlling judicial decisions take precedence over legislative summaries, government press releases, and secondary commentary. A provision that has been invalidated or whose enforcement has been suspended nationwide by a binding court order in effect on December 31, 2027, will not count as being legally operative for this question. A provision challenged in court but not subject to such an order will continue to qualify if its statutory requirements are otherwise satisfied. The resolution date is December 31, 2027, using the legal status of qualifying federal legislation at the end of that calendar day in Washington, D.C. Information published after that date may be used to establish what the law was on the resolution date.
Fine print
The question concerns federal statutory preemption of state AI-specific regulation, not the broader constitutional principle that federal law takes precedence over conflicting state law. Preemption must be express rather than merely inferred from a potential conflict between federal and state regulatory requirements. Partial preemption qualifies when it restricts a substantial, identifiable class of state AI-specific requirements. For example, a federal law barring states from imposing their own licensing requirements on developers of frontier AI models would qualify. A law preempting only one narrowly defined use of AI, without broader implications for state AI regulation, would not. A temporary moratorium qualifies if it is legally effective on December 31, 2027, even if it will expire shortly afterward. Conversely, legislation enacted during 2027 with a relevant effective date in 2028 does not qualify. A statute conditioning federal funding on states agreeing not to enforce AI regulations will qualify only if it legally prohibits enforcement of a substantial category of state AI laws by states subject to the condition. Merely offering financial incentives to avoid AI regulation, without imposing a binding restriction, does not qualify. State laws concerning privacy, consumer protection, employment, discrimination, or other general subjects are relevant only insofar as the federal statute expressly preempts AI-specific requirements within those areas. Ordinary federal legislation incidentally affecting AI companies does not qualify. The question does not require a court to have upheld the constitutionality of the qualifying statute, nor does it require any state AI law to have been successfully challenged under it. A statute's legally operative express preemption provision is sufficient, subject to the judicial-order exception described in the resolution criteria.
More public forecasts
Will the US Congress pass a federal law that limits state AI laws before 31 December 2027?
59%chance of yes
Prolepsia forecast
Will a UK law banning ransomware payments by public sector bodies be in force by 31 December 2027?
18%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
Will the EU adopt a single company law for startups across member states, often called EU Inc, by 31 December 2027?
81%chance of yes
Prolepsia forecast
Will a binding US federal measure restricting the public release of AI model weights be in effect before 1 January 2028?
29%chance of yes
Prolepsia forecast
Will the US 45Z clean fuel production tax credit still be in force on 31 December 2027?
99%chance of yes
Prolepsia forecast