ExecutiveOrdersEO
EO 14110Executive OrderBiden · D

Executive Order 14110

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

This executive order establishes a comprehensive, government-wide framework for governing AI development and use, directing federal agencies to develop safety standards, require reporting from companies building large AI models, protect against AI-enabled cyber and biological threats, safeguard civil rights and privacy, and build federal AI workforce capacity. It invokes the Defense Production Act and International Emergency Economic Powers Act to impose reporting obligations on AI developers and cloud infrastructure providers regarding dual-use foundation models and foreign user transactions.

Federal Register Open PDF 88 FR 75191

Impact dates

  1. APNSA and OMB coordinate mandatory critical infrastructure AI guidelines

  2. All life-sciences funding agencies require screened nucleic acid procurement

  3. NIST establishes AI safety guidelines, red-teaming standards, and testing benchmarks

  4. DOD and DHS report on AI cyber defense pilot results

  5. Commerce proposes foreign reseller identity verification regulations

  6. DHS incorporates AI Risk Management Framework into critical infrastructure guidelines

  7. DOD and DHS complete AI cyber defense operational pilot projects

  8. DHS evaluates AI CBRN misuse risks and reports to President

  9. OSTP establishes synthetic nucleic acid screening framework

  10. Treasury issues AI cybersecurity best practices report for financial institutions

  11. DOD contracts National Academies for AI biosecurity study

  12. Commerce requires DPA reporting from dual-use foundation model developers and computing cluster operators

  13. Commerce proposes IaaS provider regulations for foreign training run reporting

  14. DHS and sector agencies complete critical infrastructure AI risk assessments

Key directives

  • Commerce/NIST to establish AI safety guidelines and red-teaming standards within 270 days
  • Commerce to require ongoing reporting from dual-use foundation model developers under Defense Production Act within 90 days
  • Commerce to propose IaaS provider regulations for foreign training run reporting within 90 days
  • Commerce to propose foreign reseller identity verification regulations within 180 days
  • DHS to complete critical infrastructure AI risk assessments within 90 days, annually thereafter
  • Treasury to issue AI cybersecurity best practices report for financial institutions within 150 days
  • DHS to incorporate AI Risk Management Framework into critical infrastructure guidelines within 180 days
  • DOD and DHS to complete AI cyber defense pilot projects within 180 days, report within 270 days
  • DHS to evaluate AI CBRN misuse risks and report to President within 180 days
  • DOD to contract National Academies study on AI biosecurity risks within 120 days
  • OSTP to establish synthetic nucleic acid screening framework within 180 days
  • All life-sciences funding agencies to require screened nucleic acid procurement within 180 days of framework establishment

Who is ordered

Timeline

Immediate

  • Defense Production Act reporting requirements triggered for dual-use foundation model developers and large computing cluster operators
  • IaaS provider reporting requirements for foreign training runs
  • Critical infrastructure AI risk assessments initiated

Near term (90d)

  • NIST guidelines for AI red-teaming and dual-use foundation model safety
  • Commerce Department technical conditions for reporting thresholds
  • DHS critical infrastructure AI risk assessments due
  • Treasury report on AI cybersecurity risks for financial institutions
  • DOD and DHS operational pilot projects for AI-enabled cyber defense

Long term

  • Mandatory adoption of AI Risk Management Framework for critical infrastructure
  • Federal workforce AI training and hiring modernization
  • International AI governance framework development
  • Synthetic nucleic acid screening framework implementation across funding agencies

Risks & tensions

  • Defense Production Act invocation for peacetime AI regulation may face legal challenge from affected companies
  • Threshold-based reporting (10^26 operations) creates cliff effects and potential regulatory arbitrage as models approach thresholds
  • IaaS foreign user identification requirements may conflict with privacy laws in other jurisdictions
  • Dual-use foundation model definition's breadth ('could be easily modified') creates uncertainty for developers
  • Mandatory framework adoption for critical infrastructure depends on subsequent agency rulemaking with uncertain timeline
  • CBRN risk assessments require balancing scientific openness against security classification
  • Labor and civil rights provisions are largely aspirational with few hard deadlines or enforcement mechanisms
  • Federal AI workforce hiring faces competition with private sector compensation
Executive Order 14110: Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence · Executive Orders