Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence

Section 1 Purpose. Artificial intelligence (AI) holds extraordinary potential

Responsible AI development and use can address urgent challenges and improve prosperity, productivity, innovation, and security, but irresponsible AI use risks fraud, discrimination, bias, disinformation, worker displacement, reduced competition, and national security threats. The Administration commits to a society-wide effort involving government, private sector, academia, and civil society to govern AI development and use safely and responsibly, viewing this as urgent for American security, economy, and society.

Section 2 Policy and Principles

Federal agencies shall adhere to eight guiding principles for AI governance: AI systems must be safe and secure through robust evaluations, risk mitigation, and security measures addressing biotechnology, cybersecurity, and critical infrastructure risks with post-deployment monitoring and labeling; responsible innovation and competition must be promoted through investment in AI education, talent attraction, intellectual property protection, and preventing dominant firms from using key assets like semiconductors and computing power to disadvantage competitors; AI development must support American workers through collective bargaining and job training while protecting them from surveillance, rights violations, job quality degradation, reduced competition, health and safety risks, and harmful labor disruptions; policies must advance equity and civil rights by preventing unlawful discrimination and bias in hiring, housing, and healthcare, holding developers and deployers accountable through technical evaluations, community engagement, and rigorous regulation; and the United States will establish leadership in responsible technology deployment by developing international frameworks with allies and competitors to manage AI risks, establish shared principles for AI safety and security, and ensure AI reduces inequalities and protects human rights.

Section 3 Definitions

Section 3 defines key terms for this order including agency, artificial intelligence as machine-based systems making predictions or decisions based on human-defined objectives, AI models and systems, AI red-teaming, commercially available information, crime forecasting, critical and emerging technologies, dual-use foundation models trained on broad data with tens of billions of parameters posing serious risks through lowering CBRN weapons barriers, enabling offensive cyber operations, or permitting oversight evasion, federal law enforcement agency, floating-point operations, foreign persons and resellers, generative AI, infrastructure-as-a-service, personally identifiable information, privacy-enhancing technologies such as secure multiparty computation or differential privacy, privacy impact assessments, sector risk management agencies, self-healing networks, synthetic biology, synthetic content modified or generated by algorithms, testbeds, and watermarking as embedding information in AI outputs to verify authenticity.

Section 4 Ensuring the Safety and Security of AI Technology

Within 270 days, NIST must establish AI safety guidelines and red-teaming benchmarks; the Secretary of Energy must develop evaluation tools for nuclear, biological, chemical, critical infrastructure and energy security risks within 270 days; and companies must report dual-use foundation models and large computing clusters (over 10^26 operations, or 10^23 for biological sequences, or 100 Gbit/s interconnection with 10^20 operations per second capacity). US Infrastructure as a Service providers must prohibit foreign resellers unless they report foreign persons training large AI models with malicious cyber potential (defined as over 10^26 operations on specified clusters) and verify foreign account holders' identities within 180 days. Within 180 days, the Secretary of Homeland Security must incorporate AI risk management into critical infrastructure standards, and the Director of OSTP must establish a framework for synthetic nucleic acid sequence screening; within 240 days after those standards are complete, federal agencies must mandate these guidelines through regulatory action.

Section 5 Promoting Innovation and Competition

Multiple agencies must streamline pathways for AI talent and innovation: State and Homeland Security must expedite visa processing for AI experts within 90-180 days; Homeland Security must modernize H-1B programs and publish multilingual resources within 120 days; USPTO must issue AI and intellectual property guidance within 270 days, consult on copyright issues within 270 days or 180 days after Copyright Office study (whichever is later), and Homeland Security must develop an IP theft mitigation program within 180 days; NSF must develop public-private AI partnerships; HHS must prioritize healthcare AI grants including equity and diversity; VA must host two 3-month AI competitions within 365 days; Energy must issue an AI grid report within 180 days and establish an AI coordination office; SBA must fund Regional Innovation Clusters for AI institutes and allocate up to 2 million dollars in Growth Accelerator bonuses for AI-focused accelerators.

Section 6 Supporting Workers. (a) To advance the Government’s understanding

Within 180 days, the Council of Economic Advisers shall report on AI's labor-market effects and the Secretary of Labor shall report on Federal agencies' capacity to support workers displaced by AI and identify options including legislative measures; the Secretary of Labor shall also establish within 180 days guidance for employers on AI workplace deployment addressing job displacement, labor standards, equity, compensation, health, safety, data collection, and transparency, with agencies considering promotion of these guidelines where appropriate and consistent with law. The National Science Foundation Director shall prioritize resources for AI education and workforce development through existing programs, consult with agencies about additional opportunities, and use fellowship programs and awards for these purposes.

Section 7 Advancing Equity and Civil Rights

The Attorney General shall coordinate with agencies to enforce existing Federal civil rights laws related to AI, convene within 90 days a meeting of Federal civil rights offices on algorithmic discrimination, consider providing guidance to prosecutors on investigating discrimination in automated systems, and within 365 days submit a report on AI use in criminal justice covering sentencing, parole, bail, risk assessments, police surveillance, crime forecasting, prison management, and forensic analysis with recommendations on safeguards and use limits. Within 180 days, the Secretary of HHS must prepare a plan on automated or algorithmic systems in public benefits administration promoting access assessment, recipient notice, evaluation for unfair denials, human discretion, human appeals, and equitable outcomes analysis; within 180 days, the Secretary of Agriculture must issue guidance on such systems to maximize access, follow merit system requirements, identify notification requirements, allow human appeal and support, enable auditing, and assess equitable outcomes; and within 365 days, the Secretary of Labor must publish guidance for federal contractors on nondiscrimination in AI-based hiring systems.

Section 8 AI in transportation and education

Independent regulatory agencies are encouraged to protect consumers from fraud, discrimination, privacy threats, and other AI risks through rulemaking and clarifying how existing regulations apply to AI. HHS must establish an AI Task Force within 90 days to develop a strategic plan within 365 days on responsible AI deployment in healthcare and human services, a quality assurance strategy within 180 days for pre- and post-market oversight, technical assistance on nondiscrimination compliance within 180 days, and an AI safety program within 365 days to track clinical errors; HHS must also develop a drug development regulation strategy within 365 days. The Secretary of Transportation must direct the NETT Council within 30 days to assess AI information needs, support pilot programs, and establish a Cross-Modal Executive Working Group; within 90 days request DOT advisory committee advice on safe AI use; and within 180 days direct ARPA-I to explore AI opportunities and challenges in transportation.

Section 9 Protecting Privacy. (a) To mitigate privacy risks potentially exacerbated

The Director of OMB must evaluate and report on commercially available information procured by agencies, particularly personally identifiable information from data brokers and vendors, and develop guidance on mitigating privacy and confidentiality risks; within 180 days issue an RFI on strengthening privacy impact assessments for AI; and within 365 days the Secretary of Commerce through NIST must create guidelines for agencies to evaluate differential-privacy protections for AI. Within 240 days, the Director of NSF must work with agencies to identify opportunities to incorporate privacy-enhancing technologies and prioritize research to encourage their adoption, informed by results from the United States-United Kingdom privacy-enhancing technologies prize challenge.

Section 10 Advancing Federal Government Use of AI

Within 60 days, the OMB Director must establish an interagency AI council led by OMB with OSTP as Vice Chair, and within 150 days issue guidance requiring agencies to designate a Chief AI Officer within 60 days, implement minimum risk-management practices for AI affecting rights or safety, and identify specific high-impact AI uses. Within 60 days OMB must develop AI adoption tracking, within 90 days NIST must develop minimum risk-management guidelines, and within 180 days OMB must align agency AI contracts with guidance and establish means to coordinate AI procurement. OMB must issue annual instructions on collecting and publishing agency AI use cases; agencies should limit specific generative AI services based on risk assessments rather than broad bans.

Section 11 Strengthening American Leadership Abroad. (a) To strengthen United

The Secretary of State shall lead international engagement to expand collaboration and establish frameworks managing AI risks, while the Secretary of Commerce shall develop a global standards plan within 270 days covering AI terminology, data protection, trustworthiness, and risk management guided by the NIST AI Risk Management Framework. Within 365 days, the Secretary of State and USAID Administrator shall publish a playbook for AI in international development incorporating NIST principles. The Secretaries of State and Energy, NSF Director, and USAID Administrator must create a plan for worldwide AI research promoting safe development and addressing labor consequences; DHS must coordinate with State and partners on preventing critical infrastructure disruptions from AI, develop a multilateral engagement proposal within 270 days supporting adoption of AI safety guidelines for infrastructure operators, and report to the President within 180 days thereafter on priority steps to reduce cross-border threats to American infrastructure.

Section 12 White House Artificial Intelligence Council

Section 13 General provisions

The order does not alter legal authorities of executive departments or agencies or the OMB Director's budgetary functions. Implementation must comply with applicable law and appropriations availability. The order creates no enforceable legal rights or benefits against the United States, its departments, agencies, officers, employees, or other parties.