The Trump administration's AI doomer moment

Casey Newton · rss · 2026-05-05 source ↗ Reader 🖥️

AI Backlash AI Regulation & Governance (External)AI-Driven Vulnerability Discovery Mythos Project Glasswing Software Supply-Chain Security

The Trump administration, having spent over a year dismissing AI safety concerns as fearmongering and regulatory capture, has been forced by the demonstrable national-security risks of Anthropic's Mythos model to quietly revive the very pre-release government review regime it mocked and dismantled under Biden — revealing that accelerationist AI policy collapses when confronted with actual frontier-model capabilities.

practical · political analysis / opinion column — tracks the reversal of the Trump administration's AI policy from deregulatory accelerationism to safety-conscious oversight — Genre: tech-policy journalism with editorial commentary. Narrates a policy U-turn and interprets its causes and likely consequences.

Structure

Part	Role
The accelerationist thesis (Vance, Sacks, Kratsios)	Establishes the administration's prior ideological position: AI safety is overblown, regulation kills innovation, and safety advocates are engaged in regulatory capture.
The anti-regulation campaign (state-level lobbying)	Shows the administration actively working to block state AI regulations, demonstrating the position was not merely rhetorical but operationally pursued.
The pivot: pre-release government review (NYT report)	Presents the core news: Trump is reviving Biden-era pre-release model review via executive order and a working group — the very mechanism his officials ridiculed.
The catalyst: Mythos and national security	Identifies what forced the change — Anthropic's Mythos model demonstrated cybersecurity exploit capabilities that the government deems a national security risk.
The Anthropic contradictions	Exposes the incoherence: one part of the administration designated Anthropic a 'supply chain risk' while another part is expanding Mythos access throughout government; officials are simultaneously trying to block and promote the same technology.
Industry compliance and the renamed Safety Institute	Shows the outcome looks functionally identical to Biden's regime — Google, Microsoft, xAI submitting models for review, handled by the renamed AI Safety Institute.
Forward-looking consequences	Surveys likely effects: state regulation moratorium unlikely to pass, possible tighter export controls, potential for abuse of licensing power by a partisan administration.

Key terms

doomer industrial complex AI Backlash
David Sacks's pejorative label for AI safety advocates, researchers, and organizations whom he accused of fearmongering to achieve regulatory capture.
Mythos Mythos
Anthropic's latest frontier LLM, currently in limited preview, whose cybersecurity exploit capabilities triggered the administration's policy reversal.
pre-release government review AI Regulation & Governance (External)
A regulatory mechanism requiring AI companies to submit frontier models for government safety evaluation before wide release — the Biden-era approach that Trump revoked and is now reviving.
supply chain risk Software Supply-Chain Security
A legal designation the Trump administration applied to Anthropic after it refused to amend its Pentagon contract to allow 'all lawful use' — used as a punitive tool that now conflicts with the same administration's desire to use Anthropic's technology.
Center for AI Standards and Innovation AI Regulation & Governance (External)
The renamed US AI Safety Institute under Commerce, which will handle the pre-release model reviews — functionally the same body the administration rebranded to signal its anti-safety stance.

Propositions

explicitcorroboratedThe Trump administration is reviving Biden-era pre-release government review of frontier AI models, a policy its own officials previously ridiculed and dismantled.
supportsCould generative AI turn out to be the tech industry's Vietnam? And could public backlash lead AI to a better path? · p5 — The administration reviving Biden-era pre-release review it previously ridiculed is a confirming instance of Marcus's prediction that Trump will distance himself from aggressively pro-AI policy.
supports ←The AI safety vibe shift · p4 — Both document the same broader September 2026 political re-engagement with AI safety risk, from different institutional angles (Congressional legislation vs. administration policy reversal).
explicitcontestedAnthropic's Mythos model has demonstrated cybersecurity exploit capabilities that the government considers a national security risk.
contradicts ←NHS England rushes to hide software over AI hacking fears · p4 — UK AISI's evaluation found Mythos only capable of attacking small, weakly defended systems — undercutting the national-security-risk framing of its cyber capabilities.
explicitThe White House opposes Anthropic's plan to expand Mythos access from ~50 to 120 companies for security reasons.
explicitGoogle, Microsoft, and xAI have agreed to give the government early access to their models for evaluation.
explicitThe administration is simultaneously trying to block Anthropic as a 'supply chain risk' and expand government access to Anthropic's technology — an internally contradictory position.
explicitThe administration's effort to place a moratorium on state-level AI regulation now seems less likely to pass.
explicitcorroboratedPublic backlash against data centers and AI infrastructure is putting the administration at odds with its own political base.
supports ←AI's environmental costs threaten water, land and climate · p8 — Substantiates the political-backlash claim with concrete data-centre resource burdens (electricity share, water draws amid drought) plausibly driving the local opposition described.
implicitIdeological commitments to deregulation are brittle — they collapse when abstract policy meets concrete capability demonstrations.
refinesSebastian Mallaby on AI Safety and the Race for Superintelligence · p10 — Adds a third path to Mallaby's disaster-or-governance dichotomy: a concrete capability demonstration short of disaster (Mythos) forced governance.
refinesCould generative AI turn out to be the tech industry's Vietnam? And could public backlash lead AI to a better path? · p8 — Locates the mechanism of policy reversal in concrete capability demonstrations (Mythos) rather than Marcus's diffuse cultural backlash.
implicitThe Biden administration's AI safety framework was substantively correct in identifying the need for pre-release review, even if politically unpopular with the incoming administration.
implicitcorroboratedThe speed of AI capability advancement outpaced the political cycle, forcing a mid-course correction the administration did not anticipate needing.
supports ←The AI safety vibe shift · p3 — Both attribute the shift in political/public attention to AI capability advancing faster than institutions and public opinion could track.
implicitdevelopedAnthropic's refusal to grant blanket 'all lawful use' terms to the Pentagon was principled rather than obstructionist, as the very capabilities it sought to gate are now recognized as genuinely dangerous.
extends ←Walter Isaacson finally got scared about AI · p6 — Both describe the same Anthropic-vs-Pentagon dynamic (Anthropic declining blanket 'all lawful use' terms); this document adds the claim that OpenAI stepped in to fill the gap, extending the 'principled restraint' framing with a competitive-erosion mechanism the other document doesn't cover.
implicitRenaming the AI Safety Institute to 'Center for AI Standards and Innovation' was cosmetic — the underlying institutional function was needed and has now been restored in practice.

Arguments

The administration's accelerationist AI policy has failed on its own terms. strength: high

Explicit premises:

The administration's position was that AI safety fears were overblown fearmongering.
A concrete frontier model (Mythos) has demonstrated capabilities that even the administration acknowledges pose national security risks.
The administration is now reviving the exact pre-release review mechanism it dismantled.

Implicit premises:

If your policy was correct, you would not need to adopt the policy of your opponents.

Inferences:

The reversal itself is the evidence of failure — the administration's actions refute its own prior rhetoric.

The Trump administration holds an internally contradictory position on Anthropic. strength: high

Explicit premises:

One set of officials designated Anthropic a 'supply chain risk' and is working to phase out its models over six months.
Another set of officials is working to expand agency access to Anthropic's Mythos model throughout government.

Implicit premises:

A coherent policy cannot simultaneously exclude and depend on the same supplier.

Inferences:

The contradiction reveals that the supply chain risk designation was punitive rather than analytical — it was about the Pentagon contract dispute, not actual risk assessment.

Pre-release government review of AI models is becoming the bipartisan default regardless of which party holds power. strength: medium

Explicit premises:

Biden introduced pre-release review via executive order.
Trump revoked it, then revived a functionally equivalent mechanism.
All major AI companies (Google, Microsoft, xAI) have already agreed to comply.

Implicit premises:

When both opposing administrations converge on the same policy, the policy is driven by structural necessity rather than ideology.

Inferences:

Pre-release review is likely a durable feature of AI governance regardless of future political shifts.

Narratives

The administration began with confident dismissal of AI safety ('hand-wringing'), actively dismantled Biden-era safeguards, lobbied against state regulations, and then was forced by Mythos's capabilities to reverse course and adopt the very framework it had mocked — a classical hubris-to-humility arc.
→ Provides the central dramatic structure of the column: overconfidence meets reality, making the policy reversal legible as a story rather than just a policy change.
Despite political hostility, the institutional infrastructure for AI safety (the Safety Institute, pre-release review processes, export controls) proved more durable than the political rhetoric against it — renamed and rhetorically disowned but not destroyed, now quietly re-empowered.
→ Suggests that safety institutions have a structural resilience that survives political cycles, supporting the implicit argument that pre-release review is a bipartisan default.

Questions addressed

✅ What caused the Trump administration to reverse its anti-safety AI policy?
Anthropic's Mythos model demonstrated cybersecurity exploit capabilities that pose national security risks, forcing the administration to acknowledge that safety concerns were substantive.
✅ How does the new pre-release review regime compare to Biden's?
Functionally similar — companies submit models for government evaluation before wide release, handled by the same institution (renamed). The mechanism is nearly identical.
✅ What is the administration's position on Anthropic, and why is it contradictory?
One faction designated Anthropic a supply chain risk; another is expanding Mythos access throughout government. The contradiction stems from the punitive designation colliding with genuine need for the technology.
⭕ What are the likely downstream effects of this reversal on state regulation and export controls?
State regulation moratorium less likely to pass; export controls may tighten with Sacks's departure. But outcomes remain uncertain.
⭕ What risks does a Trump-administered AI licensing regime pose?
Newton raises the possibility of censorship — denying releases to 'woke' models or pressuring companies politically — but does not assess how likely this is.
⭕ Do other frontier models besides Mythos pose similar national security risks?
Not addressed — the article focuses exclusively on Mythos as the catalyst.

Gaps

missing evidenceThe article does not detail what specific cybersecurity exploits Mythos demonstrated, making it hard to assess the proportionality of the government's reaction.
Read next: National security reporting on AI-enabled cyber capabilities; Anthropic's own responsible scaling disclosures.
closed byNHS England rushes to hide software over AI hacking fears — Reports the UK AI Security Institute's evaluation of Mythos — capable only of attacking small, weakly defended, vulnerable enterprise systems — supplying the capability detail needed to assess the proportionality of the government's reaction.
unasked questionNo discussion of whether other frontier models (from Google, OpenAI, xAI) have similar capabilities — the narrative is anchored entirely on Anthropic/Mythos, suggesting the policy shift may be broader than one model.
Read next: Comparative assessments of frontier model capabilities across labs; government red-team evaluations.
unasked questionThe piece does not explore whether the 'supply chain risk' designation against Anthropic was legally sound or politically motivated beyond noting the Pentagon contract dispute.
Read next: Legal analysis of the supply chain risk designation framework and its application to AI companies.
scope limitationNo analysis of how foreign governments (China, Russia) are responding to these same capability thresholds — the national security framing is entirely domestic.
Read next: International AI governance developments; Chinese AI regulation and military AI capabilities research.
implicit premiseThe article assumes the pre-release review mechanism will function effectively but does not examine the Commerce Department's actual capacity to evaluate frontier models — a critical institutional capacity question.
Read next: NIST/Commerce Department AI evaluation capabilities; UK AISI's experience with model evaluation.
tensionCasey Newton discloses that his fiancee works at Anthropic but does not explore how this shapes the article's Anthropic-centric framing — the piece treats Anthropic as the protagonist of the policy reversal.
Read next: Coverage of the same events from non-Anthropic-adjacent journalists for comparison.

Closes gaps elsewhere

AI Companies Aren't Evil. But They Are Reckless. · g3 — Documents the speed mismatch the essay omits: AI capability advancement outpaced the political cycle, forcing a mid-course governance correction the administration did not anticipate.
Could generative AI turn out to be the tech industry's Vietnam? And could public backlash lead AI to a better path? · g1 — Supplies an empirical instance of the missing backlash-to-policy transmission: data-center backlash from the administration's own base plus fear of blowback from AI-enabled cyberattacks produced an actual policy reversal.

The Trump administration's AI doomer moment

Structure

Key terms

Propositions

Arguments

Narratives

Questions addressed

Gaps

Closes gaps elsewhere

Related documents