The 2025 pivot: two orders, two philosophies

Lesson 1 of 5 in The Federal Posture: Executive Orders, OMB, and the Preemption Fight.

Start with the structural fact that explains everything else in this module: the United States still has no comprehensive federal AI statute. In that vacuum, federal AI policy is made by executive orders and OMB memoranda — instruments that bind federal agencies (and, through procurement, their contractors), but generally not private companies, and that a new president can revoke with a signature.

That fragility is not a footnote. It is the defining property of the US federal posture, and January 2025 demonstrated it in the most dramatic way possible.

President Biden’s EO 14110 (30 October 2023) was the most ambitious AI directive any US administration had issued — roughly a hundred pages tasking dozens of agencies. Its sharpest edge used the Defense Production Act to compel developers of dual-use foundation models above a compute threshold to report safety-test results to the government — the first mandatory obligation on frontier developers anywhere in US law. It also ordered the NIST guidance that became the Generative AI Profile, stood up the US AI Safety Institute, and pushed agencies toward rights-protective deployment practices.

On 20 January 2025 — inauguration day — EO 14148 revoked it wholesale. Three days later, EO 14179, ‘Removing Barriers to American Leadership in AI’ (23 January 2025), installed the replacement philosophy: sustain US dominance, develop AI systems ‘free from ideological bias or engineered social agendas’, review and unwind EO 14110-era actions, and deliver an AI Action Plan within 180 days. As of September 2026, EO 14179 remains the foundation of federal AI policy.

EO 14110 regime (Oct 2023 – Jan 2025)

Organizing idea: safe, secure, and trustworthy AI. Risk and rights sit alongside innovation.

  • Mandatory frontier reporting — Defense Production Act authority forced developers of dual-use foundation models above compute thresholds to share red-team and safety-test results with government.
  • Agency tasking at scale — dozens of deadlines across Commerce, DHS, HHS, Labor, DOJ; content-provenance work; guidance on discrimination in housing, hiring, and benefits.
  • US AI Safety Institute at NIST; the Generative AI Profile (NIST-AI-600-1) commissioned.
  • Companion soft law — sat atop the 2022 Blueprint for an AI Bill of Rights and OMB M-24-10/M-24-18, which built ‘safety-impacting’ and ‘rights-impacting’ AI categories for agencies.

EO 14179 regime (Jan 2025 – present)

Organizing idea: American AI dominance. Regulation is framed as the risk; the barrier to remove.

  • Revocation and review — EO 14110 rescinded (via EO 14148); agencies ordered to identify and unwind actions taken under it.
  • ‘Free from ideological bias’ — a new policy criterion that later drives LLM-procurement rules (EO 14319) and the directive to revise the NIST AI RMF.
  • AI Action Plan in 180 days — delivered 23 July 2025 with roughly 90 policy actions across three pillars.
  • Institutional renaming with substance — the AI Safety Institute becomes CAISI (Center for AI Standards and Innovation, June 2025), its mission tilted toward security, standards leadership, and competitiveness rather than ‘safety’ framing.
  • OMB M-25-21/M-25-22 (April 2025) replace the Biden memos — Chief AI Officers and inventories survive; the two risk categories collapse into one ‘high-impact’ tier with an innovation-first tone.

What survived the pivot

Revocation is blunter than reality. Working through what survived is the best calibration exercise in US AI policy:

  • The NIST Generative AI Profile (July 2024) — born from EO 14110, still current and cited after the revocation. Technical artifacts outlive their political parents.
  • Chief AI Officers, use-case inventories, minimum practices — rebranded and re-scoped in M-25-21, not abolished.
  • The AI RMF itself — untouched as the reference framework, though now under revision on the new administration’s terms.
  • Agency enforcement authority — the FTC’s Section 5, ECOA, Title VII never depended on any EO (next module).

What died: the DPA reporting mandate as policy centerpiece, the rights-impacting framing, and the Blueprint for an AI Bill of Rights as an operative reference.

Executive-order genealogy, 2019–2026

  1. EO 13859 (2019)

    Maintaining American Leadership in AI — Trump’s first-term order: R&D investment, NIST tasked with technical standards leadership. The seed of NIST’s central role.

  2. EO 14110 (Oct 2023)

    Safe, Secure, and Trustworthy AI — DPA frontier reporting, AI Safety Institute, GenAI Profile commissioned. REVOKED Jan 20, 2025.

  3. EO 14148 (Jan 20, 2025)

    Initial Rescissions — day-one order revoking EO 14110 among dozens of Biden actions.

  4. EO 14179 (Jan 23, 2025)

    Removing Barriers to American Leadership in AI — innovation-first policy, ideological-bias language, 180-day Action Plan directive. IN FORCE.

  5. AI Action Plan (Jul 23, 2025)

    America’s AI Action Plan — ~90 actions, three pillars; NIST RMF revision directed.

  6. EO 14319 + data-center + export EOs (Jul 23, 2025)

    Companion orders: Preventing Woke AI (federal LLM procurement principles), accelerated data-center permitting, promotion of US AI-stack exports. IN FORCE.

  7. EO 14365 (Dec 11, 2025)

    Ensuring a National Policy Framework for AI — DOJ directed to challenge state AI laws; federal funding conditioned; carve-outs for child-safety, data-center, and procurement laws. IN FORCE.

  8. EO 14409 (Jun 2026)

    Promoting Advanced AI Innovation and Security — reported voluntary frontier-model framework with NSA-supported benchmarking. Single-source details; verify before relying.

Key terms: executive order, omb, caisi, soft law, preemption

Interactive checkpoint quiz (2 questions) — open this page in a browser to take it.