Superseded: Rescinded by Trump Executive Order 14148 on January 20, 2025, and replaced conceptually by Executive Order 14365 — Ensuring a National Policy Framework for AI and America's AI Action Plan. Retained here for historical reference.
Executive Order 14110, "Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence," was signed on October 30, 2023 by President Joseph R. Biden, Jr. and published at 88 FR 75191 (Nov 1, 2023; document 2023-24283). It was an omnibus AI directive spanning more than 100 specific agency taskings across safety, civil rights, privacy, labor, immigration, healthcare, competition, and international engagement, and was the longest AI-focused executive order signed as of 2023. It was rescinded by Trump Executive Order 14148 on January 20, 2025.
Key provisions
The order's frontier-safety core was a dual-use foundation model reporting requirement (§ 4.2). Acting through the Defense Production Act, the Commerce Department required developers training models above a set compute threshold to report training runs, physical and cybersecurity protections, model-weight ownership, and red-team results. The threshold was operationalized as 10²⁶ FLOPs for dense models, with a lower threshold for sequence-specific biology models.
Under § 4.1, the order directed NIST to develop evaluations for dual-use foundation models, red-teaming guidance, and a companion Generative AI Profile, published as NIST AI 600-1. A related provision (§ 4.2(c)) required US Infrastructure-as-a-Service providers to collect identifying information on foreign customers training large AI models, a measure described as IaaS "Know Your Customer." On synthetic content provenance, § 4.5 directed Commerce to develop watermarking and content-authentication standards. On biosecurity, § 4.4 directed HHS to address AI-enabled CBRN risks, including screening of synthetic nucleic-acid providers.
Beyond frontier safety, the order addressed immigration (§ 5), streamlining visas for AI talent and expanding Schedule A occupations; civil rights (§ 7), requiring agencies to issue guidance on AI-enabled discrimination in housing, lending, employment, and healthcare; federal use of AI (§ 10), establishing Chief AI Officers at each agency, an inventory of federal AI uses, and generative-AI pilots; and international engagement (§ 11), promoting global common approaches that informed the later The Bletchley Declaration (AI Safety Summit, 1–2 November 2023) and the G7 Hiroshima process.
Key definitions
The order defined a dual-use foundation model as one trained on broad data, self-supervised, with at least tens of billions of parameters, and capable or easily modifiable to enable CBRN design, offensive cyber operations, or evasion of human control. It defined AI red-teaming as structured adversarial testing for flaws, vulnerabilities, and misuse, and generative AI as a class of models emulating input structure to produce synthetic content.
Rescission and residual effect
Trump EO 14148 rescinded EO 14110 within hours of inauguration on January 20, 2025. Several downstream work products persisted past the rescission. NIST AI 600-1 (NIST AI 600-1 — Generative AI Profile) was not withdrawn. The US AI Safety Institute was restructured but remained functional. The BIS compute-reporting rules continued in modified form (see BIS Framework for AI Diffusion — Interim Final Rule (Jan 13, 2025)). Agency Chief AI Officers were retained under EO 14365. The successor framework is described in Executive Order 14365 — Ensuring a National Policy Framework for AI (Trump, December 2025) and America's AI Action Plan.
Two underlying claims of the order have been treated as contestable. The premise that Commerce holds statutory authority under the Defense Production Act to compel compute-based AI reporting has been contested in some legal commentary and was never fully litigated (confidence: medium). The premise that a 10²⁶ FLOPs threshold identifies the frontier was rated high at the time of the order but has become increasingly contested as compute-efficiency gains make smaller models capable (confidence: high at time of order).
Relationships
- supersedes: prior AI policy baseline established by Executive Order 13859 — Maintaining American Leadership in Artificial Intelligence (Trump 2019)
- superseded-by: Executive Order 14365 — Ensuring a National Policy Framework for AI (Trump 2025), which rescinded EO 14110 via EO 14148
- depends-on: NIST AI Risk Management Framework (AI RMF 1.0) — EO 14110 built on and commissioned expansion of the RMF
- supports: NIST AI 600-1 — Generative AI Profile, BIS Framework for AI Diffusion — Interim Final Rule (Jan 13, 2025) — commissioned under this EO
- related: Statement on AI Risk (CAIS), FLI — Pause Giant AI Experiments: An Open Letter — advocacy backdrop
- related: California SB 1047 — Safe and Secure Innovation for Frontier AI Models Act (enrolled + veto) — state-level parallel using the same 10²⁶ FLOPs threshold
- contradicts: America's AI Action Plan — successor framework reversing the safety-reporting posture