Federal Trustworthy AI Executive Order

Overview

Executive Order 13960, Promoting the Use of Trustworthy Artificial Intelligence in the Federal Government, was signed by President Donald J. Trump on December 3, 2020. It was published on December 8, 2020, as Federal Register Document 2020-27065, at 85 FR 78939–78943. Issued by the President of the United States, the instrument establishes principles and administrative mechanisms for trustworthy AI use by federal executive departments and agencies. It is an executive order rather than a statute enacted by Congress or a general licensing system for private AI developers.

The order encourages appropriate adoption of AI to improve public services and government operations while protecting privacy, civil rights, civil liberties, and public trust. Its stated purpose addresses federal uses for purposes other than national security and defense. Its implementation structure combines nine substantive principles with common Office of Management and Budget guidance, inventories of agency uses, assessment of operational applications, corrective or retirement planning, and information sharing. These measures connect the benefits of AI adoption with ongoing responsibility for its performance and effects.

Definitions and Scope

The order addresses the design, development, acquisition, and use of AI applications by covered federal agencies. Its principal subjects include executive departments and agencies and the officials and personnel responsible for those activities. The term Principles refers to the nine requirements enumerated in section 3. Those principles apply across the AI lifecycle rather than only when an agency first purchases or deploys a system. Inventory requirements also encompass current and planned uses consistent with agency missions.

Section 5 requires inventories of non-classified and non-sensitive AI use cases within the scope defined by section 9. An AI use-case inventory is an agency catalogue of such covered uses, prepared initially and annually thereafter. NIST's implementation page describes its inventory as covering non-classified, non-sensitive, and non-research uses. That page also reports no operational use cases meeting the relevant scope and discusses reevaluation in fiscal year 2023. This agency-specific implementation statement concerns NIST's inventory rather than the inventories of the federal government as a whole.

Governance and Institutional Framework

The Office of Management and Budget has the central policy-coordination role. Section 2(c) provides for common OMB policy guidance governing implementation of the principles, and section 4 directs OMB to publicly post a roadmap identifying AI-related guidance it intends to create or revise. Existing OMB policies concerning information and information technology continue to apply where consistent with the principles and applicable law. In preparing its roadmap, OMB must consider its broader guidance instruments, including circulars and management memoranda, and include public-engagement arrangements and timelines for final guidance where appropriate.

The Federal Chief Information Officers Council establishes and publishes inventory criteria, formats, and mechanisms. Individual agencies prepare inventories and assess their applications, while agency-designated responsible officials approve plans to achieve consistency or retire inconsistent applications. Implementation involves the Agency Data Governance Body and relevant officials outside that body. Interagency inventory sharing is coordinated through the CIO Council, the Chief Data Officer Council, and other appropriate bodies. This structure places AI governance within established federal policy, information-management, and agency decision-making arrangements.

Trustworthy AI Principles

Section 3 requires agencies to adhere to nine principles when designing, developing, acquiring, and using AI. Applications must be lawful and respectful of national values; purposeful and performance-driven; accurate, reliable, and effective; safe, secure, and resilient; understandable; responsible and traceable; regularly monitored; transparent; and accountable. Lawfulness includes consistency with the Constitution and applicable laws and policies, including protections for privacy, civil rights, and civil liberties. Purposeful use means seeking opportunities where benefits significantly outweigh risks and where the risks can be assessed and managed.

Accuracy, reliability, and effectiveness require consistency with the use cases for which applications were trained. Safety, security, and resilience encompass systematic vulnerabilities, adversarial manipulation, and malicious exploitation. Understandability is calibrated to subject-matter experts, users, and others as appropriate. Responsibility and traceability require clearly assigned human responsibilities and appropriate, practicable documentation of AI development, acquisition, use, and relevant inputs and outputs. Together, the principles address operational quality, protection of rights, human responsibility, and the ability to examine how an application is used.

Policy Guidance and Implementation

Section 4(b) requires OMB to publicly post its policy-guidance roadmap within 180 days of the order. The roadmap identifies guidance to be developed or revised in support of trustworthy AI use and provides an administrative path for addressing novel aspects of AI within existing federal information and technology policies. Section 4(c) directs agencies to continue using available voluntary consensus standards developed with industry participation where their use is consistent with law and not impracticable. OMB must also consider those standards when developing its guidance.

Section 5(a) gives the Federal CIO Council 60 days from the order to identify and publish the criteria, format, and mechanisms for agency inventories. Under section 5(b), agencies must prepare inventories within 180 days after completion of the Council's directive and annually thereafter. This deadline is tied to completion of the inventory directive rather than directly to the signing date. Including planned as well as current uses makes the inventory a mechanism for prospective visibility alongside documentation of existing agency activity.

Assessment, Monitoring, and Corrective Action

Section 5(c) requires agencies to identify, review, and assess existing operational AI applications for inconsistencies with the order. Within 120 days of completing their inventories, agencies must develop and obtain responsible-official approval for plans to achieve consistency or retire inconsistent applications. Section 5(c)(ii) directs agencies to strive to implement approved plans within 180 days of approval, subject to existing resource levels. The planning and approval deadline is distinct from the qualified implementation direction, and the plans may require coordination across data-governance and other agency functions.

Regular monitoring is also an independent principle under section 3(g). Agencies must ensure that AI applications are regularly tested against the principles and maintain mechanisms to supersede, disengage, or deactivate applications whose performance or outcomes conflict with their intended use or the order. Accountability under section 3(i) includes appropriate safeguards, compliance monitoring, auditing, documentation, and personnel training. Annual inventories and ongoing testing serve different functions: an inventory identifies uses, while monitoring evaluates whether applications continue to operate consistently with their purpose and the governing principles.

Transparency, Accountability, and Protected Interests

The transparency principle calls for appropriate disclosure to stakeholders, including Congress and the public. Disclosure is qualified by practicability and applicable laws protecting privacy and sensitive information. Sections 5(d) and 5(e) add specific inventory-sharing requirements: agencies must share inventories with other agencies within 60 days of inventory completion and make inventories publicly available within 120 days, subject to practicability and applicable information-protection requirements. These measures support oversight and learning while retaining protections for information that should not be publicly released.

The order links accountability to human responsibility, documentation, training, monitoring, and audit safeguards. It directs agency conduct through operational controls, including assessment, remediation, retirement, and intervention when an application's performance or outcomes are inconsistent with its intended use. Lawfulness remains a separate requirement: federal AI use must comply with the Constitution and applicable laws and policies protecting privacy, civil rights, and civil liberties. These protections accompany the order's performance and governance requirements rather than being displaced by the expected benefits of an AI application.

Relationship to Other Instruments

Section 1 aligns the order with Executive Order 13859, Maintaining American Leadership in Artificial Intelligence, signed on February 11, 2019. It also identifies the Department of Defense's AI Ethical Principles dated February 24, 2020, and the intelligence community's AI ethics principles and framework dated July 23, 2020. These instruments form part of the surrounding United States policy landscape and explain the distinction between the order's stated purpose and existing national-security and defense arrangements.

Section 4 preserves existing OMB information and information-technology policies where they are consistent with the order's principles and applicable law. The order therefore works alongside established federal management rules rather than automatically replacing them. A later federal AI instrument is Executive Order 14110, Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, signed on October 30, 2023. The identification of this later instrument provides chronological policy context without assigning it a specific legal effect on Executive Order 13960.

Standards and Federal Policy Orientation

The order's standards-related mechanism is the continued use of available voluntary consensus standards developed with industry participation where lawful and practicable. This gives agencies a route to use shared technical and organizational practices while retaining their responsibility to apply the nine principles. OMB's obligation to consider those standards connects agency implementation with common policy development. The direction is conditional: availability, consistency with law, and practicability matter when agencies decide how standards fit a particular application or operating environment.

The instrument's substantive orientation is federal-government governance. Reliability, transparency, accountability, protection of rights, and risk management are expressed as requirements for federal AI activities. Its references to defense and intelligence ethics instruments also concern United States institutions. Standards-informed implementation supports consistency across agencies, while agency-specific inventories and assessments accommodate different missions and operational uses. The resulting framework combines common principles with distributed execution through existing federal bodies and responsible officials.

Implementation Timeline

Compliance Checklist

Sources and References