The Information Machine

US AI Regulation: Federal Retreat vs. State Intervention · history

Version 18

2026-06-15 02:48 UTC · 294 items

What

American AI governance runs on three parallel tracks: a Trump executive order establishing voluntary 30-day pre-release review with NSA oversight [1][2]; enacted and advancing state mandates in Illinois and California [8][9][10]; and Florida's tort and criminal actions against OpenAI [13]. Anthropic CEO Dario Amodei published a major blog calling for mandatory pre-release testing and government-backed independent auditing of frontier models [16], while framing AI as the likely dominant source of military and economic power for any nation [17]. David Sacks — who stripped binding requirements from the Trump EO — is now publicly critical of Anthropic [28], bringing the conflict between the accelerationist faction and labs that advocate mandatory oversight into the open.

Why it matters

Amodei's call for mandatory (not voluntary) pre-release testing puts Anthropic further from the current federal framework than any major lab has publicly gone. Sacks' public turn against Anthropic suggests the accelerationist faction now views labs advocating mandatory oversight as adversaries. The federal apparatus remains voluntary and of uncertain enforceability, leaving state law as the only source of binding AI requirements.

Open questions

  • Will Sacks' public criticism of Anthropic [28] translate into executive-branch pressure on the company, or remain a public dispute without regulatory consequence?

  • Can Amodei's call for mandatory pre-release testing and government-backed auditing [16] gain traction when the accelerationist faction demonstrably shapes the current executive AI framework [6][7]?

  • Does Amodei acknowledging that 'terrible mistakes may be made' from military use of Claude while maintaining usage red lines [18] prove commercially sustainable, or does it become a liability with government clients?

  • Does Build American AI's admission that it hired an outside vendor to run sockpuppet accounts [24] produce FEC or legal consequences for the PAC or its backers?

Narrative

Trump signed a revised AI executive order on June 3 establishing a voluntary framework for pre-release review of frontier models, with a 30-day government access window and NSA oversight [1][2]. The order explicitly disclaims creating a licensing regime or binding requirements on AI firms [3]. Analyst Zvi Mowshowitz argues it functions as de facto mandatory because labs have no practical choice but to participate, and notes that NSA holds the oversight role with classified benchmarking thresholds that prevent AI lab researchers from knowing whether their own models trigger review [1][4]. DOGE-driven cuts to federal cybersecurity and safety teams have reduced the government's practical capacity to conduct substantive evaluations within the 30-day window [5]. The order's weakened form reflects successful lobbying by Musk, Zuckerberg, and Sacks, who stripped binding requirements and reduced the original review window from 90 to 30 days [6][7].

At the state level, Illinois enacted SB 315 as the nation's most stringent AI safety statute, requiring major AI firms to publish annual third-party safety test results, report critical incidents within 72 hours, and providing whistleblower protections [8]. California is advancing SB 947 requiring human oversight of automated employment decisions [9] and SB 951 requiring 90-day advance notice of AI-driven layoffs [10]. The Trump administration's preemption executive order challenged state AI laws on constitutional grounds [11], but multiple law firms found its authority limited absent congressional authorization [12]. Florida AG Uthmeier filed the first state-led lawsuit against OpenAI and Sam Altman, arguing the company marketed ChatGPT as safe while knowing it posed harm to children, and separately opened a criminal investigation [13][14][15].

AnthropicCEO Dario Amodei published a blog in June 2026 calling for mandatory pre-release testing of frontier AI models before deployment and independent auditing backed by government authority, arguing that AI capability progress has outpaced governments' ability to regulate it [16]. Amodei frames the geopolitical stakes plainly: powerful AI will become the dominant source of military and economic power for any nation, equivalent to having 'a country of geniuses in a datacenter' [17]. In a separate interview, Amodei acknowledged that 'terrible mistakes' may result from military uses of Claude while defending Anthropic's practice of maintaining red lines around permissible uses — even when doing so risks commercial viability [18]. Anthropic also published a public opinion survey showing 70% of Americans across party lines support government AI regulation, while only 15% trust AI companies to make governance decisions — the lowest of any institution tested [19].

OpenAI has produced a coordinated set of public policy documents calling for federal oversight over fragmented state regulation [20][21][22], while its affiliated PAC Build American AI — backed by OpenAI, Andreessen/a16z, and Palantir [23] — admitted hiring an outside vendor to run anonymous sockpuppet accounts targeting opponents [24][25][26] and separately paid TikTok influencers to fear-monger about Chinese AI [27]. David Sacks, who has served as the administration's primary public defender of AI against 'fear-based regulation' [28] and successfully stripped binding requirements from the Trump EO [6][7], is now publicly critical of Anthropic — placing the conflict between the accelerationist faction and labs that advocate mandatory oversight explicitly on the record.

Timeline

  • 2025-12: Trump signs executive order aimed at eliminating state law obstruction of national AI policy [29][30][49]
  • 2026-01-01: New state AI laws take effect nationally, triggering the federal preemption debate [50]
  • 2026-05: Trump signs preemption EO challenging state AI laws; multiple law firms find constitutional authority limited absent congressional authorization [11][30][12][47][48]
  • 2026-05: California SB 951 advances requiring 90-day advance notice of AI-driven layoffs, backed by California Federation of Labor [10][36]
  • 2026-05-21: Trump cancels AI security EO after Sacks calls Trump directly to oppose pre-deployment review; Musk and Zuckerberg also lobby against it [6][7][51]
  • 2026-05: California Senate approves SB 947 requiring human oversight of automated decision systems in employer discipline and termination [9][37]
  • 2026-05-28: Illinois Governor Pritzker signs SB 315 — nation's strictest AI safety statute, requiring mandatory third-party auditing, 72-hour critical incident reporting, and whistleblower protections [38][39][8]
  • 2026-05-28: OpenAI publishes Frontier Governance Framework supporting thoughtful AI regulation [22][52]
  • 2026-06-01: Florida becomes first state to sue OpenAI and Sam Altman over ChatGPT's alleged role in violent crimes; AG Uthmeier also opens a separate criminal investigation [14][15][13]
  • 2026-06-02: Senator Sanders announces the American AI Sovereign Wealth Fund Act, proposing a one-time 50% stock transfer from major AI companies to a public sovereign fund [44][45]
  • 2026-06-03: Trump signs revised AI executive order establishing a voluntary 30-day pre-release review for frontier models with NSA oversight and classified thresholds [1][2][4][3]
  • 2026-06-03: OpenAI publishes a public policy agenda calling for federal framework over fragmented state regulation [20][32][33]
  • 2026-06-04: Build American AI — backed by OpenAI, Andreessen/a16z, and Palantir — acknowledged running false flag social media accounts; WIRED separately reports the PAC paid TikTok influencers to fear-monger about Chinese AI [25][23][26][27]
  • 2026-06-08: Jensen Huang refuses Senator Warren's request to testify before Congress on Nvidia AI chip exports to China and US export control policy [46]
  • 2026-06-09: OpenAI publishes Industrial Policy paper advocating for people-first AI-era governance focused on expanding opportunity and sharing AI-generated prosperity [21]
  • 2026-06-10: Anthropic CEO Dario Amodei publishes blog calling for mandatory pre-release testing of frontier AI models and independent auditing backed by government authority, framing AI as the likely dominant source of future military and economic power [16][17]
  • 2026-06-12: Anthropic publishes the first Public Record survey: 70% of Americans support government AI regulation across party lines; only 15% trust AI companies to make governance decisions — lowest of any institution tested [19]
  • 2026-06-12: Amodei acknowledges in interview that 'terrible mistakes' may result from military use of Claude while defending Anthropic's practice of maintaining red lines on permissible uses [18]
  • 2026-06-14: David Sacks publicly turns critical of Anthropic, marking an open break between the accelerationist faction and labs that advocate mandatory AI oversight [28]

Perspectives

Trump administration (federal)

Signed a voluntary AI EO with NSA oversight and a 30-day review window; explicitly disclaims creating a licensing regime or binding requirements on AI firms.

Evolution: Signed a significantly weakened version of the original order after accelerationist lobbying stripped binding requirements and reduced the review window from 90 to 30 days.

[29][11][30][6][1][2][4][3]

Tech accelerationist faction (Musk, Zuckerberg, Sacks)

Actively opposes AI oversight mechanisms; Sacks is now publicly critical of Anthropic specifically, placing the faction in open conflict with labs that advocate mandatory oversight.

Evolution: Consistent on opposing regulation; Sacks' public turn against Anthropic is new — the conflict with labs favoring mandatory oversight is now explicit rather than structural.

[6][7][1][31][28]

Anthropic / Dario Amodei

Calls for mandatory pre-release testing and independent government-backed auditing of frontier AI; frames AI as potentially the dominant source of military and economic power; acknowledges military use risks while maintaining red lines on permissible uses; published public opinion survey showing only 15% of Americans trust AI companies on governance.

Evolution: Significantly expanded: Amodei's blog and interview add an explicit mandatory-oversight position and a geopolitical framing that go further than prior Anthropic statements, and draw a public response from Sacks.

[19][18][17][16]

OpenAI

Publicly advocates for federal governance and equitable industrial policy while its affiliated PAC admitted hiring an outside vendor to run sockpuppet accounts targeting opponents and paid TikTok influencers to fear-monger about Chinese AI.

Evolution: The vendor admission confirms and extends the prior false flag disclosure; the gap between OpenAI's published advocacy and its PAC's documented tactics is now fully on the record.

[22][20][32][33][25][23][26][27][34][21][35][24]

State governors advancing AI legislation (Newsom + Pritzker)

State government must proactively address AI harms through legislation; Illinois SB 315 is enacted as the nation's strictest AI safety statute; California advances SB 947 and SB 951.

Evolution: Consistent; the two-state legislative coalition continues operating independently of federal coordination.

[10][36][9][37][38][39][8]

Florida AG Uthmeier (litigation track)

Pursues AI accountability through civil tort and criminal investigation; filed the first state-led lawsuit against OpenAI arguing it marketed ChatGPT as safe while knowing it posed harm to children.

Evolution: Consistent; litigation and criminal investigation proceed independently of the state legislative and federal preemption debates.

[14][15][13][40][41]

Congress — federal legislative responses

Activity is fragmented: the PREPARE Act frames AI workforce concerns as development opportunities; Sanders proposed a 50% stock transfer to a sovereign fund; Senator Warren pressed Nvidia on chip exports to China but Jensen Huang refused to testify.

Evolution: Consistent; congressional activity remains fragmented with no binding legislation advancing.

[42][43][44][45][31][46]

Zvi Mowshowitz (independent analyst)

The Trump AI EO is de facto mandatory despite voluntary framing, with NSA oversight and classified thresholds preventing labs from knowing their own compliance status; DOGE cuts leave the government unable to conduct substantive reviews; OpenAI's denial of the Build American AI false flag operation is not credible.

Evolution: Consistent; original reporter on the Build American AI false flag story.

[1][5][25]

Tensions

  • EO design vs. enforcement reality: The Trump AI EO is nominally voluntary, but Mowshowitz argues labs have no practical choice but to comply, NSA holds the oversight role with classified thresholds labs cannot see, and DOGE cuts leave the government unable to conduct substantive evaluations [1][4][5]. [1][2][5][4][3]
  • Mandatory oversight (Anthropic/Amodei) vs. accelerationist faction (Sacks): Amodei calls for mandatory pre-release testing and government-backed auditing [16]; Sacks stripped binding requirements from the Trump EO [6][7] and is now publicly critical of Anthropic [28]. [6][7][16][28]
  • Federal preemption vs. state governance: The Trump preemption EO claims to block state AI laws, but multiple law firms find its constitutional authority limited absent congressional authorization, potentially leaving state laws intact until courts rule [11][12]. [11][30][12][47][48]
  • OpenAI's governance advocacy vs. its PAC's political operations: OpenAI publicly calls for democratic, equitable governance [20][21] while Build American AI admitted hiring a vendor to run sockpuppet accounts [24][25] and paid TikTok influencers to spread fear about Chinese AI [27]; only 15% of Americans trust AI companies on governance [19]. [20][21][25][26][27][19][24]
  • OpenAI's federal framework preference vs. the state-driven landscape: OpenAI advocates for coherent federal architecture to displace fragmented state rules [32][33], while binding laws from Illinois and California advance without federal coordination [8][9]. [20][32][33][38][8]
  • Voluntary federal framework vs. Amodei's mandatory oversight call: The Trump EO explicitly disclaims binding requirements [3] while Amodei calls for mandatory pre-release testing and government-backed independent auditing [16] — two incompatible regulatory postures for the same moment. [3][16]

Sources

  1. [1] Trump Signs Executive Order For AI Testing Prior To Frontier Model Releases — Zvi's AI Roundups (2026-06-03)
  2. [2] President Trump signs executive order to review advanced AI models. — Rohan Paul Twitter (2026-06-03)
  3. [3] Executive Order Creates Voluntary Regulatory Regime of Frontier AI ... — reactive:us-ai-policy-regulation
  4. [4] Trump AI Executive Order Gives NSA Classified Role Over Frontier Models — reactive:us-ai-policy-regulation
  5. [5] Trump plan to test AI models has a problem—US security teams were gutted by DOGE — Ars Technica AI (2026-06-03)
  6. [6] Trump abruptly cancels EO signing event after top AI firm CEOs declined to go — Ars Technica AI (2026-05-22)
  7. [7] AI #170: Lack of Executive Order — Zvi's AI Roundups (2026-05-28)
  8. [8] Illinois Signs Frontier AI Safety Bill into Law | Chris Lehane posted ... — reactive:us-ai-policy-regulation
  9. [9] CA Senate Approves No Robo Bosses Act of 2026 to Ensure Human Oversight of AI in the Workplace | Senator Jerry McNerney — reactive:us-ai-policy-regulation
  10. [10] Bill Text: CA SB951 | 2025-2026 | Regular Session | Amended — reactive:us-ai-policy-regulation
  11. [11] President Trump Signs Executive Order Preempting State AI Laws ... — reactive:us-ai-policy-regulation
  12. [12] President Trump Signs Executive Order Challenging State AI Laws | Paul Hastings LLP — reactive:us-ai-policy-regulation
  13. [13] Florida AG brings 'first-in-the-nation state-led lawsuit' against OpenAI and its CEO — reactive:us-ai-policy-regulation
  14. [14] ChatGPT creators knew product would cause harm, Florida argues in lawsuit • Florida Phoenix — reactive:us-ai-policy-regulation
  15. [15] Attorney General James Uthmeier Launches Criminal Investigation into OpenAI, ChatGPT | My Florida Legal — reactive:us-ai-policy-regulation
  16. [16] Dario Amodei just published a super long blog, calling for an urgent policy overhaul because he thinks frontier AI is mo… — Rohan Paul Twitter (2026-06-10)
  17. [17] Dario Amodei argues in his new blog today, that powerful AI will soon become the dominant source of military and economi… — Rohan Paul Twitter (2026-06-11)
  18. [18] Anthropic's Dario Amodei's new interview: on U.S. military use of Claude. — Rohan Paul Twitter (2026-06-12)
  19. [19] Results from the first Anthropic Public Record — Anthropic News (2026-06-12)
  20. [20] OpenAI public policy agenda — OpenAI Blog (2026-06-03)
  21. [21] Industrial policy for the Intelligence Age — OpenAI Blog (2026-06-09)
  22. [22] OpenAI’s Frontier Governance Framework — OpenAI Blog (2026-05-28)
  23. [23] The OpenAI-Andreessen-Palantir SuperPAC admits that it was “part ... — reactive:us-ai-policy-regulation
  24. [24] Build American AI admits hiring an outside vendor to run anonymous sockpuppet accounts targeting its opponents · Digg — reactive:us-ai-policy-regulation
  25. [25] AI #171: False Flag — Zvi's AI Roundups (2026-06-04)
  26. [26] A Pro-AI Super PAC’s Secret Meme Sockpuppets — reactive:us-ai-policy-regulation
  27. [27] A Dark-Money Campaign Is Paying Influencers to Frame Chinese AI as a Threat | WIRED — reactive:us-ai-policy-regulation
  28. [28] David Sacks is done being polite about Anthropic (Save this). — Milk Road AI Twitter (2026-06-14)
  29. [29] Ensuring a National Policy Framework for Artificial Intelligence — reactive:us-ai-policy-regulation
  30. [30] President Trump Signs Executive Order to Block State AI Laws — reactive:us-ai-policy-regulation
  31. [31] 💯💯 David Sacks rips Bernie Sanders' 50% AI ownership plan https://t.co/hPsHZ6eOyJ https://t.co/yD9Ko0V6oP — Rohan Paul Twitter (2026-06-07)
  32. [32] A blueprint for democratic governance of frontier AI — OpenAI Blog (2026-06-03)
  33. [33] 🟡 Let’s get physical — Semafor Technology (2026-06-03)
  34. [34] Introducing the OpenAI Economic Research Exchange — OpenAI Blog (2026-06-08)
  35. [35] OpenAI Says Fake Accounts From China Tried To Turn Americans Against Data Centers — reactive:us-ai-policy-regulation
  36. [36] SB 951: Employment: technological displacement: notice. — reactive:us-ai-policy-regulation
  37. [37] California Set to Restrict AI Use in the Workplace With “No Robo ... — reactive:us-ai-policy-regulation
  38. [38] Trump loses more control over AI regulation as Illinois passes landmark law — Ars Technica AI (2026-05-28)
  39. [39] Illinois Legislature passes historic AI bill that would require third-party safety audits — reactive:us-ai-policy-regulation
  40. [40] Florida Sues OpenAI Over Chatbot Safety Concerns — reactive:us-ai-policy-regulation
  41. [41] Florida AG Sues OpenAI, Says ChatGPT Is Aiding Violence — reactive:us-ai-policy-regulation
  42. [42] Text - S.3339 - 119th Congress (2025-2026): AI Workforce PREPARE Act — reactive:us-ai-policy-regulation
  43. [43] Rep. Obernolte, Rep. Jacobs Introduce Bipartisan Bill to Prepare American Workers for AI-Driven Economic Change | Representative Jay Obernolte — reactive:us-ai-policy-regulation
  44. [44] Sanders Sovereign Wealth Fund Plan Would Give US Public 'Direct Ownership Stake' in AI Giants | Common Dreams — reactive:us-ai-policy-regulation
  45. [45] Sen. Bernie Sanders on X: "AI is built on humanity’s collective knowledge. The wealth it generates must benefit humanity — not just Elon Musk, Sam Altman and other AI oligarchs. That’s why I’ll be introducing the American AI Sovereign Wealth Fund Act — to give the public a direct ownership stake. https://t.co/UqW71FBv2Z" / X — reactive:us-ai-policy-regulation
  46. [46] Jensen Huang just refused Sen. Elizabeth Warren’s request to testify before Congress on AI chips, China sales, and expor… — Rohan Paul Twitter (2026-06-08)
  47. [47] When Federal Preemption Meets AI Regulation: What Trump's Draft Executive Order Means for Your Compliance Strategy | Jones Walker LLP — reactive:us-ai-policy-regulation
  48. [48] Executive Preemption and the Dormant Commerce Clause After ... — reactive:us-ai-policy-regulation
  49. [49] Executive Order: Ensuring a National Policy Framework for Artificial Intelligence (Donald Trump, 2025) - Ballotpedia — reactive:us-ai-policy-regulation
  50. [50] New State AI Laws are Effective on January 1, 2026, But a New ... — reactive:us-ai-policy-regulation
  51. [51] WATCH: Trump explains why he postponed signing AI executive order — reactive:us-ai-policy-regulation
  52. [52] Our views on AI policy and political advocacy — OpenAI Blog (2026-06-01)