AI Wiki

Executive Order 14110

AI HistoryAI Policy & RegulationAI Safety
17 min read
Updated
Suggest editHistoryTalk26 citationsFact-checked
RawGraph

Executive Order 14110, formally titled "Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence," was signed by President Joe Biden on October 30, 2023 and published in the Federal Register on November 1, 2023 at 88 FR 75191 [1]. It ran 36 Federal Register pages and set a government-wide agenda for artificial intelligence. The Congressional Research Service counted more than 100 specific actions assigned to over 50 federal entities across eight policy areas: safety and security, innovation and competition, worker support, AI bias and civil rights, consumer protection, privacy, federal use of AI, and international leadership [2].

Section 4.2 is the part that drew the most attention. It invoked the Defense Production Act to require companies to report to the federal government on any model trained with more than 10^26 integer or floating-point operations [1]. That provision, together with a lower 10^23 threshold for models trained primarily on biological sequence data, tied federal oversight to training compute rather than to the application or sector in which a model was used.

It did not survive the change of administration. President Donald Trump revoked EO 14110 on January 20, 2025 through Executive Order 14148, "Initial Rescissions of Harmful Executive Orders and Actions" [3]. Three days later, Executive Order 14179, "Removing Barriers to American Leadership in Artificial Intelligence," replaced its policy framing and ordered a government-wide review of everything done under it [4]. Most of the reporting machinery EO 14110 set in motion had reached the proposed-rule stage and no further.

Background

EO 14110 built on two earlier, non-binding documents: the Office of Science and Technology Policy's Blueprint for an AI Bill of Rights, published in October 2022, and the NIST AI Risk Management Framework (NIST AI 100-1), released in January 2023 [2]. Neither imposed obligations on anyone. The executive order kept the same conceptual vocabulary but attached deadlines to it, most of them counted from the date of signature, with 90, 120, 180, and 270 days the intervals that recur most often [1].

An executive order binds federal agencies, not private companies. Where EO 14110 reached beyond government, it did so by directing agencies to use statutory authority they already held. The reporting requirements rested on the Defense Production Act (50 U.S.C. 4501 et seq.); the cloud computing provisions rested on the national emergency declared in Executive Order 13694 of April 1, 2015, as amended by Executive Orders 13757 and 13984 [1]. That structure explains why little of the order ever bound anyone outside government: each hook required a rulemaking, and neither of the two that mattered was finished.

Section 3 supplied the definitions that later regimes borrowed. A "dual-use foundation model" was an AI model trained on broad data, generally using self-supervision, containing at least tens of billions of parameters, applicable across a wide range of contexts, and exhibiting or easily modified to exhibit high performance at tasks posing serious risk to security, including substantially lowering the barrier of entry for non-experts to design chemical, biological, radiological, or nuclear weapons, enabling offensive cyber operations through automated vulnerability discovery, or permitting evasion of human control through deception or obfuscation. The definition explicitly covered models shipped with safeguards intended to block those capabilities [1]. "AI red-teaming" was defined as a structured testing effort to find flaws and vulnerabilities in an AI system, often in a controlled environment and in collaboration with developers [1].

Reporting under the Defense Production Act

Section 4.2(a) gave the Secretary of Commerce 90 days to require two classes of report. Companies developing or intending to develop potential dual-use foundation models had to provide ongoing information on planned training activities and the physical and cybersecurity protections around them, on ownership and possession of model weights and the measures protecting them, and on the results of red-teaming based on guidance NIST was to write. Until that guidance existed, companies were to report red-team results on four specific capabilities: lowering the barrier to biological weapons development by non-state actors, discovering software vulnerabilities and developing exploits, using software or tools to influence real or virtual events, and self-replication or propagation [1]. Separately, anyone acquiring, developing, or possessing a large-scale computing cluster had to report its existence, location, and total computing power.

Section 4.2(b) directed Commerce to define the technical conditions that would trigger these duties and to update them over time. Until it did, interim thresholds applied.

TriggerInterim technical condition (EO 14110 Section 4.2(b))
Model training runTrained using more than 10^26 integer or floating-point operations
Biological sequence modelTrained primarily on biological sequence data using more than 10^23 integer or floating-point operations
Computing clusterMachines physically co-located in a single datacenter, transitively connected by data center networking of over 100 Gbit/s, with a theoretical maximum computing capacity of 10^20 integer or floating-point operations per second for training AI

Sections 4.2(c) and 4.2(d) addressed cloud providers. Within 90 days, Commerce was to propose rules requiring United States Infrastructure as a Service providers to report when a foreign person used their services for a training run with potential malicious cyber capability, defined on an interim basis as more than 10^26 operations run on a qualifying cluster. Within 180 days, it was to propose know-your-customer rules obliging foreign resellers of those services to verify the identity of foreign account holders [1].

The Bureau of Industry and Security published the know-your-customer proposal on January 29, 2024 at 89 FR 5698 [5]. It published the model and cluster reporting proposal on September 11, 2024 at 89 FR 73612, with comments due October 11, 2024 [6]. The second proposal would have required quarterly notifications by email and floated a revised cluster definition using networking greater than 300 Gbit/s rather than 100 Gbit/s, with maximum performance greater than 10^20 operations per second for AI training measured without sparsity. Each rulemaking identifier, RIN 0694-AJ35 and RIN 0694-AJ55, is associated with a single Federal Register document, and neither reached a final rule before the order was revoked [5][6].

Standards, testing, and the AI Safety Institute

Section 4.1 gave NIST 270 days to produce guidelines for developing and deploying safe AI systems, including a generative AI companion to the AI Risk Management Framework, a companion to the Secure Software Development Framework, an initiative on benchmarks for evaluating AI capabilities with a focus on cybersecurity and biosecurity, and guidelines enabling developers of dual-use foundation models to conduct red-teaming [1]. NIST delivered on July 26, 2024, exactly 270 days after signature: final versions of NIST AI 600-1 (the Generative AI Profile), NIST SP 800-218A, and NIST AI 100-5 (a plan for global engagement on AI standards), plus a draft of NIST AI 800-1 on managing misuse risk for dual-use foundation models and Dioptra, an open-source testbed for evaluating adversarial robustness [7].

The order did not itself create an AI Safety Institute. Commerce announced the U.S. Artificial Intelligence Safety Institute at NIST on November 1, 2023, two days after signature, to carry out the department's assignments under the order; the announcement coincided with Secretary Gina Raimondo's attendance at the AI Safety Summit in the United Kingdom [8]. Elizabeth Kelly was named inaugural director on February 7, 2024, with Elham Tabassi as chief technology officer [9]. On August 29, 2024 the institute signed memoranda of understanding with Anthropic and OpenAI providing access to major new models before and after public release for collaborative safety research [10].

Section 4.8 directed a National Security Memorandum on AI. It was issued on October 24, 2024 and designated the AI Safety Institute as the primary United States government point of contact with private developers for voluntary pre- and post-deployment testing of frontier models [11].

Biosecurity, synthetic content, and open weights

Section 4.4 addressed chemical, biological, radiological, and nuclear (CBRN) misuse, stating a particular focus on biological weapons. It gave the Department of Homeland Security 180 days to evaluate the misuse potential of AI for CBRN development and report to the President, and gave the Department of Defense 120 days to contract with the National Academies for a study of AI and biosecurity risk. It also directed OSTP to establish, within 180 days, a framework encouraging providers of synthetic nucleic acid sequences to implement procurement screening, and made adherence to that framework a condition of federal life-sciences research funding [1].

Section 4.5 covered synthetic media. Commerce had 240 days to report on existing standards for authenticating content and tracking provenance, labeling synthetic content through techniques such as watermarking, detecting synthetic content, preventing generative systems from producing child sexual abuse material or non-consensual intimate imagery, and auditing synthetic content. Commerce was then to develop guidance on tools and practices for digital content authentication and synthetic content detection, after which OMB was to instruct agencies on labeling and authenticating the digital content the government itself produces or publishes [1]. The order stopped short of requiring private labeling; it addressed federal content and federal procurement, with the Federal Acquisition Regulatory Council asked to consider amending the Federal Acquisition Regulation.

Section 4.6 handled open weights by asking a question rather than answering it. NTIA ran a public consultation, receiving 332 comments to a February 2024 request, and published "Dual-Use Foundation Models with Widely Available Model Weights" in July 2024. Its conclusion was that current evidence was not sufficient to determine either that restrictions on open-weight models were warranted or that restrictions would never be appropriate, and it recommended that the government monitor a portfolio of risks and build the capacity to act if they materialized [12].

Immigration and federal use of AI

Section 5.1 dealt with AI talent and immigration. It gave State and Homeland Security 90 days to streamline visa processing for people working in AI and other critical and emerging technologies, and 120 days for State to consider updating the 2009 Revised Exchange Visitor Skills List and implementing a domestic visa renewal program under 22 CFR 41.111(b). Within 180 days, State was to consider extending domestic renewal to academic J-1 research scholars and F-1 students in STEM fields, and Homeland Security was to clarify pathways for O-1A and EB-1 noncitizens of extraordinary ability, EB-2 advanced-degree holders, and startup founders using the International Entrepreneur Rule, while continuing its rulemaking to modernize the H-1B program [1].

Section 10.1 governed how agencies themselves use AI, and it produced the parts of the order that outlived it. OMB had 60 days to convene an interagency council on federal AI use, chaired by the OMB Director with the OSTP Director as vice chair, and 150 days to issue government-wide guidance. That guidance was OMB Memorandum M-24-10, "Advancing Governance, Innovation, and Risk Management for Agency Use of Artificial Intelligence," issued March 28, 2024 by Director Shalanda D. Young. It required each agency to designate a Chief AI Officer within 60 days, established internal AI Governance Boards, and set minimum risk-management practices for safety-impacting and rights-impacting AI that agencies had to meet by December 1, 2024 or stop using the system [13]. A companion acquisition memorandum, M-24-18, followed on September 24, 2024 [14]. Section 12 created a White House Artificial Intelligence Council chaired by the Assistant to the President and Deputy Chief of Staff for Policy [1].

Revocation

Executive Order 14148, signed January 20, 2025 and published at 90 FR 8237, listed 78 Biden-era executive actions for revocation, 67 of them executive orders. EO 14110 appears as item (ggg) [3]. The rescission carried no explanation specific to AI; the order's stated purpose was to remove what it described as inflationary and radical practices across the federal government.

Executive Order 14179, signed January 23, 2025 and published at 90 FR 8741, supplied the reasoning and the replacement. It declared it the policy of the United States "to sustain and enhance America's global AI dominance in order to promote human flourishing, economic competitiveness, and national security." Section 5(a) directed the Assistant to the President for Science and Technology, the Special Advisor for AI and Crypto, and the Assistant to the President for National Security Affairs to review every policy, directive, regulation, and order issued under "the revoked Executive Order 14110" and to suspend, revise, or rescind anything inconsistent with the new policy. Section 5(b) gave OMB 60 days to revise M-24-10 and M-24-18. Section 4 ordered an AI action plan within 180 days [4].

Replacement policy, 2025 and 2026

The successor framework arrived in stages.

DateActionEffect on EO 14110's legacy
April 3, 2025OMB M-25-21 and M-25-22, signed by Director Russell T. VoughtRescind and replace M-24-10 and M-24-18 [15][16]
June 3, 2025Commerce statement by Secretary Howard LutnickReforms the U.S. AI Safety Institute into the Center for AI Standards and Innovation (CAISI) [17]
July 23, 2025America's AI Action PlanOver 90 federal policy actions across innovation, infrastructure, and international diplomacy and security [18][19]
July 23, 2025Executive Orders 14318, 14319, 14320Data center permitting, ideological neutrality conditions on federal procurement of large language models, and promotion of American AI technology stack exports [25]
December 11, 2025Executive Order 14365, "Ensuring a National Policy Framework for Artificial Intelligence"Targets state AI regulation and calls for a single national standard [20]
June 2, 2026Executive Order 14409, "Promoting Advanced Artificial Intelligence Innovation and Security"Voluntary federal access to "covered frontier models" before release [21]

The AI Action Plan describes EO 14110 as an order "that foreshadowed an onerous regulatory regime" and treats its rescission as a completed deregulatory step [18]. CAISI kept much of the evaluation work its predecessor had started, including assessments of foreign open-weight models and joint exercises with the United Kingdom's institute, while reframing the mission around security, standards, and international competition rather than safety [17][24].

EO 14409 is the closest the later framework comes to reproducing what EO 14110 attempted, and the difference is in the compulsion. It directs a classified benchmarking process, keyed to the advanced cyber capabilities of AI models, to set the threshold at which a model is designated a "covered frontier model," and asks agencies to design a framework through which developers could give the government access to such models for up to 30 days before they plan to release them to other trusted partners. Participation is voluntary, and the order states that nothing in that section authorizes a mandatory governmental licensing, preclearance, or permitting requirement for developing, publishing, releasing, or distributing new AI models [21].

Legacy

EO 14110's substantive footprint outlasted its legal life mainly through its numbers, and the borrowing shows up most clearly in California. SB 1047, the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, defined a covered model to include one trained with more than 10^26 integer or floating-point operations at a cost above $100 million, and defined a computing cluster with the same 100 Gbit/s networking and 10^20 operations per second figures the executive order had used. Governor Gavin Newsom vetoed it on September 29, 2024 [22][26]. A year later, to the day, he signed SB 53, the Transparency in Frontier Artificial Intelligence Act, which defines a frontier model as "a foundation model that was trained using a quantity of computing power greater than 10^26 integer or floating-point operations" [23]. The phrasing is lifted almost verbatim from Section 4.2(b).

The federal governance layer also persisted in modified form. Chief AI Officers, agency AI governance structures, and use case inventories survived the transition from M-24-10 to M-25-21, with the risk-management vocabulary rewritten around "high-impact AI" rather than safety-impacting and rights-impacting categories [15]. What did not survive was the mandatory element: the two Bureau of Industry and Security rulemakings that would have turned the order's reporting duties into enforceable regulation never advanced past the proposal stage [5][6], and as of mid-2026 no comparable federal reporting mandate has replaced them.

See also

References

  1. ^Executive Order 14110 of October 30, 2023, "Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence," Federal Register vol. 88, no. 210, pp. 75191-75226, November 1, 2023. govinfo.gov/...2023-24283.pdf
  2. ^Congressional Research Service, "Highlights of the 2023 Executive Order on Artificial Intelligence for Congress," R47843, November 17, 2023. congress.gov/...R47843.2.pdf
  3. ^Executive Order 14148 of January 20, 2025, "Initial Rescissions of Harmful Executive Orders and Actions," 90 FR 8237, January 28, 2025. govinfo.gov/...2025-01901.pdf
  4. ^Executive Order 14179 of January 23, 2025, "Removing Barriers to American Leadership in Artificial Intelligence," 90 FR 8741, January 31, 2025. govinfo.gov/...2025-02172.pdf
  5. ^Bureau of Industry and Security, "Taking Additional Steps To Address the National Emergency With Respect to Significant Malicious Cyber-Enabled Activities," proposed rule, RIN 0694-AJ35, 89 FR 5698, January 29, 2024. govinfo.gov/...5698
  6. ^Bureau of Industry and Security, "Establishment of Reporting Requirements for the Development of Advanced Artificial Intelligence Models and Computing Clusters," proposed rule, Docket No. 240905-0231, RIN 0694-AJ55, 89 FR 73612, September 11, 2024. govinfo.gov/...73612
  7. ^NIST, "Department of Commerce Announces New Guidance, Tools 270 Days Following President Biden's Executive Order on AI," July 26, 2024. nist.gov/...-new-guidance-tools-270-days-following
  8. ^U.S. Department of Commerce, "At the Direction of President Biden, Department of Commerce to Establish U.S. Artificial Intelligence Safety Institute to Lead Efforts on AI Safety," November 1, 2023. web.archive.org/...ommerce-establish-us-artificial
  9. ^NIST, "U.S. Commerce Secretary Gina Raimondo Announces Key Executive Leadership at U.S. AI Safety Institute," February 7, 2024. nist.gov/...-announces-key-executive-leadership-us
  10. ^NIST, "U.S. AI Safety Institute Signs Agreements Regarding AI Safety Research, Testing and Evaluation With Anthropic and OpenAI," August 29, 2024. nist.gov/...greements-regarding-ai-safety-research
  11. ^The White House, "Memorandum on Advancing the United States' Leadership in Artificial Intelligence; Harnessing Artificial Intelligence to Fulfill National Security Objectives; and Fostering the Safety, Security, and Trustworthiness of Artificial Intelligence," October 24, 2024. bidenwhitehouse.archives.gov/...he-safety-security
  12. ^National Telecommunications and Information Administration, "Dual-Use Foundation Models with Widely Available Model Weights," July 2024. ntia.gov/...ntia-ai-open-model-report.pdf
  13. ^Office of Management and Budget, Memorandum M-24-10, "Advancing Governance, Innovation, and Risk Management for Agency Use of Artificial Intelligence," March 28, 2024. bidenwhitehouse.archives.gov/...l-Intelligence.pdf
  14. ^Office of Management and Budget, Memorandum M-24-18, "Advancing the Responsible Acquisition of Artificial Intelligence in Government," September 24, 2024. bidenwhitehouse.archives.gov/...ion-Memorandum.pdf
  15. ^Office of Management and Budget, Memorandum M-25-21, "Accelerating Federal Use of AI through Innovation, Governance, and Public Trust," April 3, 2025. whitehouse.gov/...-Governance-and-Public-Trust.pdf
  16. ^Office of Management and Budget, Memorandum M-25-22, "Driving Efficient Acquisition of Artificial Intelligence in Government," April 3, 2025. whitehouse.gov/...l-Intelligence-in-Government.pdf
  17. ^U.S. Department of Commerce, "Statement from U.S. Secretary of Commerce Howard Lutnick on Transforming the U.S. AI Safety Institute into the Pro-Innovation, Pro-Science U.S. Center for AI Standards and Innovation," June 3, 2025. web.archive.org/...ward-lutnick-transforming-us-ai
  18. ^The White House, "Winning the AI Race: America's AI Action Plan," July 2025. whitehouse.gov/...Americas-AI-Action-Plan.pdf
  19. ^The White House, "White House Unveils America's AI Action Plan," July 23, 2025. whitehouse.gov/...-unveils-americas-ai-action-plan
  20. ^Executive Order 14365 of December 11, 2025, "Ensuring a National Policy Framework for Artificial Intelligence," 90 FR 58499, December 16, 2025. govinfo.gov/...58499
  21. ^Executive Order 14409 of June 2, 2026, "Promoting Advanced Artificial Intelligence Innovation and Security," 91 FR 34565, June 5, 2026. govinfo.gov/...34565
  22. ^California Senate Bill 1047 (Wiener), Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, 2023-2024 session, vetoed September 29, 2024. leginfo.legislature.ca.gov/...billTextClient.xhtml
  23. ^California Senate Bill 53 (Wiener), Chapter 138, Statutes of 2025, Transparency in Frontier Artificial Intelligence Act, approved September 29, 2025. leginfo.legislature.ca.gov/...billTextClient.xhtml
  24. ^NIST, "Center for AI Standards and Innovation (CAISI)." nist.gov/caisi
  25. ^Executive Orders 14318, 14319, and 14320 of July 23, 2025, 90 FR 35385, 90 FR 35389, and 90 FR 35393, July 28, 2025. govinfo.gov/...35385
  26. ^Office of the Governor of California, veto message returning Senate Bill 1047, September 29, 2024. gov.ca.gov/...SB-1047-Veto-Message.pdf

Improve this article

Add missing citations, update stale details, or suggest a clearer explanation. Every suggestion is reviewed for sourcing before it goes live.

v1 · 3,303 words · full history

Fact-checks are independent of edits: a reviewer re-verifies the article against its sources and stamps the date. How we verify

Research and drafting on this wiki are AI-assisted, under named human editorial standards. How AI is used here

Reviewer note: Independent adversarial fact-check at creation (wanted175 campaign, 2026-07-24): every claim verified against primary sources by a dedicated verification agent; corrections applied before publication.

Cite this page: AI Wiki. "Executive Order 14110." aiwiki.ai, updated 24 Jul 2026, fact-checked 24 Jul 2026. CC BY 4.0. https://aiwiki.ai/wiki/executive_order_14110

Suggest edit