Congress floated three-year state AI preemption June 4, Colorado replaced its AI law May 14, and the EU AI Act enforcement deadline arrives in 51 days — three accountability claims this week

Three institutions made accountability claims in the last month at the moment federal preemption meets state displacement, comprehensive frameworks meet transparency-only regimes, and EU enforcement deadlines meet legislative delay negotiations.

Representatives Jay Obernolte (R-CA) and Lori Trahan (D-MA) released a 269-page discussion draft June 4 titled the Great American Artificial Intelligence Act of 2026, proposing to freeze state laws "specifically regulating the development" of AI models for three years. Colorado Governor Jared Polis signed SB 189 on May 14, replacing the nation's most comprehensive AI law—originally scheduled to take effect June 30, 2026—with a narrower transparency framework effective January 1, 2027. And the EU AI Act's obligations for high-risk AI systems take legal effect August 2, 2026, in 51 days, though a provisional Digital Omnibus agreement reached May 7 would defer stand-alone Annex III compliance to December 2, 2027, pending formal adoption.

All three landed within the last month. All three involve operators making claims about federal displacement, state framework replacement, or enforcement timelines that can be graded against what Congress actually enacts, what Colorado actually enforces, and what the EU actually requires six to eighteen months from now.

3 Claims

Claim 1 — Representatives Obernolte and Trahan: Released discussion draft June 4, 2026, proposing three-year preemption of state laws regulating AI model development

Representatives Jay Obernolte (R-CA) and Lori Trahan (D-MA) unveiled a discussion draft on June 4, 2026, that includes a three-year preemption of state laws related to AI development . The bill would preempt state laws and regulation "specifically regulating the development" of an AI model, with a three-year sunset, though preemption would not apply to laws related to the use or deployment of AI models .

An accompanying document released by Trahan's office said that California's AB 2013 law, which requires model developers to publicly post high-level summaries of their training data, would be preempted, along with a portion of California SB 942 law related to content watermarking . The bill would require large frontier developers—those with more than $500 million in gross revenue for the previous calendar year—to establish public frontier AI frameworks, including information on whether a model could pose a "catastrophic risk" .

The claim is gradeable on whether Congress formally introduces the bill by September 12, 2026 (three months); whether any committee holds a markup or hearing by December 12, 2026 (six months); whether the administration endorses or opposes the preemption provision by August 12, 2026 (two months); and whether any state AG publicly commits to challenging federal preemption by July 12, 2026 (one month).

Grade by: 2027-06-04 (1 year)

Invalidator: If Congress enacts a federal AI framework with no preemption provision, or if courts strike down a preemption clause as violating the Tenth Amendment, the claim that this discussion draft represents a viable path to federal displacement of state AI development laws would be invalidated.

Claim 2 — Colorado Governor Jared Polis: Signed SB 189 on May 14, 2026, replacing the comprehensive Colorado AI Act with a transparency-focused framework effective January 1, 2027

On May 14, 2026, Colorado Governor Polis signed SB 189, which revises Colorado's original artificial intelligence law and delays the effective date from June 30, 2026, to January 1, 2027, while significantly scaling back its original requirements . The Act moves away from the Colorado AI Act's original risk-based framework, eliminating the duty of care aimed at preventing algorithmic discrimination, deployer obligations to maintain risk management programs and conduct impact assessments, and certain reporting obligations to the Colorado Attorney General, adopting a narrower approach focused on disclosures and transparency around certain automated decision-making technologies .

The new bill was introduced and signed within two weeks of its introduction, reflecting pressure to land the rewrite before the original AI Act's June 30, 2026 effective date and amid escalating federal headwinds . The December 2025 White House executive order specifically named Colorado's AI Act as an example of a state law that would compel AI systems to "produce false results in order to avoid a 'differential treatment or impact' on protected groups" .

The claim is gradeable on whether any company publicly announces compliance implementation by October 1, 2026 (four months before effective date); whether the Colorado Attorney General publishes implementing regulations by January 1, 2027 (the effective date); whether any entity files suit challenging SB 189 by March 1, 2027 (two months post-effective date); and whether Polis or the AG publicly state that the new framework protects Colorado consumers equivalently to the original Act by December 1, 2026.

Grade by: 2027-01-01 (effective date)

Invalidator: If the Colorado Attorney General declines to issue mandatory regulations by January 1, 2027, or if federal courts enjoin the new framework on First Amendment or Commerce Clause grounds before the effective date, the claim that Colorado has established a workable, enforceable transparency regime would be invalidated.

Claim 3 — European Union: The AI Act's high-risk system obligations take legal effect August 2, 2026, though provisional Digital Omnibus agreement would defer Annex III compliance to December 2, 2027

The AI Act entered into force on August 1, 2024, and will be fully applicable two years later on August 2, 2026, with some exceptions including that the rules for high-risk AI systems embedded into regulated products have an extended transition period until August 2, 2028 . Under the Digital Omnibus provisional trilogue agreement, stand-alone Annex III systems covering recruitment, credit scoring, law enforcement, education, and border control tools will now need to comply by December 2, 2027, and AI embedded in regulated products under Annex I by August 2, 2028, with the agreed text replacing the Commission's originally proposed conditional trigger mechanism with these fixed dates .

Formal adoption and publication in the Official Journal are expected in the coming weeks, in advance of the August 2, 2026 deadline, though these changes only take legal effect upon formal adoption and publication of the Omnibus in the Official Journal, expected before August 2, 2026 . Because trilogue negotiations have not concluded, August 2, 2026, remains the legally binding deadline for Annex III systems today, and any executive who treats the 2027 date as settled law is operating on legislative optimism rather than legal fact .

The claim is gradeable on whether the European Parliament and Council formally adopt the Digital Omnibus by August 2, 2026 (51 days); whether any member state market surveillance authority publicly confirms it will enforce Annex III obligations on August 2, 2026, absent formal Omnibus adoption; whether the European Commission publishes harmonized technical standards by June 2, 2026; and whether at least three frontier AI providers publicly commit to August 2, 2026 compliance by July 2, 2026 (one month out).

Grade by: 2026-08-02 (51 days)

Invalidator: If the Digital Omnibus is formally adopted and published in the Official Journal before August 2, 2026, deferring Annex III obligations to December 2, 2027, then the claim that August 2, 2026, represents the enforceable compliance deadline for high-risk standalone AI systems would be invalidated.

2 Reckonings

Reckoning 1 — President Trump's June 2, 2026 executive order requesting voluntary 30-day pre-release reviews: graded against public confirmation of participation

In Entry 038 (June 11, 2026), this ledger recorded President Trump's June 2, 2026 executive order asking AI developers to voluntarily submit frontier models to the government for up to 30 days before releasing them to the public. The order directed Treasury, NSA, and CISA to develop a classified benchmarking process within 60 days to assess advanced cyber capabilities and determine when a model should be designated a "covered frontier model."

The claim was gradeable on whether any frontier lab publicly confirmed providing early model access by December 2, 2026; whether the NSA, Treasury, and CISA published the benchmarking process by August 2, 2026 (60 days from order); and whether "voluntary" in practice meant labs faced implicit pressure through federal contracts, regulatory forbearance, or public statements linking cooperation to competitiveness.

What happened: Ten days have passed since the order. The order came after Anthropic announced Claude Mythos Preview, a model that excels at identifying weaknesses and security flaws within software, with the company limiting the rollout to a select group of companies as part of Project Glasswing, which it expanded on June 2 . Anthropic announced it is expanding the release of its Mythos model from roughly 50 to 200 organizations on the same day the order was signed . No frontier lab has publicly confirmed submitting a model under the 30-day voluntary framework, and the benchmarking process is due August 1, 2026—49 days out.

Grade: Incomplete (I)
The grading horizon has not yet arrived (December 2, 2026, for public confirmation; August 2, 2026, for benchmarking process). The order is ten days old. The expansion of Project Glasswing the same day the order was signed suggests Anthropic may be positioning Mythos Preview access as partial compliance with the voluntary framework, but no lab has publicly confirmed submitting a model for 30-day government review before releasing it to other trusted partners.

Invalidator: If any frontier lab publicly confirms submitting a model under the voluntary 30-day framework before public release by August 2, 2026, and that confirmation includes documentation that the lab withheld the model from all other partners (including trusted research organizations) during the government review period, the claim that the voluntary framework remained untested would be invalidated.

Reckoning 2 — Washington's digital-likeness law (SSB 5886) taking effect June 10, 2026: graded against enforcement and platform implementation

In Entry 037 (June 10, 2026), this ledger recorded that Washington's SSB 5886, protecting digital likenesses from AI deepfakes, took effect June 10, 2026, with the law enforceable by the person depicted (private right of action) and by the Washington AG, and with liability attaching to both creators and distributors with no cure period.

The claim was gradeable on whether any creator or platform was sued or prosecuted under SSB 5886 by December 10, 2026; whether platforms implemented new consent mechanisms or content moderation for AI-generated likenesses by June 30, 2026; and whether courts granted injunctions or imposed penalties under the statute by March 10, 2027.

What happened: Two days have passed since the effective date. No public filings indicate suits or prosecutions under SSB 5886 as of June 12, 2026. Major platforms (Meta, Google, X, TikTok, OpenAI) have not publicly announced Washington-specific consent mechanisms or content moderation changes in response to the effective date.

Grade: Incomplete (I)
The grading horizon has not yet arrived (December 10, 2026, for enforcement; June 30, 2026, for platform implementation). The law is two days old. The absence of immediate platform announcements is noteworthy—most state laws with platform-specific requirements trigger pre-effective-date implementation timelines or public statements of compliance posture. The silence suggests either that platforms view SSB 5886 as duplicative of existing policies, or that they are evaluating whether to challenge the statute on First Amendment grounds before implementing changes.

Invalidator: If any platform publicly announces Washington-specific AI-generated likeness consent mechanisms or content moderation policies by June 30, 2026, and cites SSB 5886 as the driver, or if the Washington AG sends demand letters or initiates enforcement actions against any platform or creator by July 10, 2026 (one month post-effective date), the claim that the law's enforcement posture remained untested would be invalidated.

1 Refusal

I refused to use "federal AI preemption" in the title without specifying that the Great American AI Act is a discussion draft, not introduced legislation, and that the preemption clause applies only to state regulation of AI model development, not deployment or use. Early coverage framed the June 4 release as "Congress moves to freeze state AI laws," which conflates a bipartisan discussion draft—released for feedback before formal introduction—with enacted or even introduced legislation. The preemption provision is three years, not indefinite, and it explicitly carves out state laws governing AI deployment, algorithmic discrimination in employment or lending, child safety, and state procurement. Framing the draft as "Congress proposes to ban state AI regulation" erases the distinction between regulating how models are built and regulating how they are used, a distinction the drafters included precisely to preserve state authority over the harms most states have actually legislated against. The claim is that Congress floated a three-year freeze on state laws regulating AI development, not that Congress moved to preempt all state AI law.

I refused to describe the Great American AI Act discussion draft as congressional action to preempt state AI regulation without specifying the three-year sunset, the development-only scope, and the carveouts for deployment, use, and consumer protection laws.

— Roger Grubb, Editor