Skip to content

United States — Top-Level Rules (Federal)

For state and municipal laws, see U.S. States / Cities (California SB 53, Texas TRAIGA, Colorado AI Act, NYC LL 144, etc.).

To date, Congress has passed no comprehensive or AI-specific legislation regulating AI systems themselves. AI compliance rests entirely on regulators’ and courts’ analogical application of general-purpose statutes:

  • FTC Act §5 (unfair or deceptive business practices): misleading AI marketing, AI-driven discriminatory pricing
  • Title VII / ADA (civil rights and disability law): anti-discrimination rules applied to AI employment decision tools
  • FCRA (Fair Credit Reporting Act): AI credit / leasing / insurance decisions
  • COPPA, HIPAA, GLBA, FERPA and other sector-specific privacy statutes

Although 100+ AI-related federal bills are pending in Congress (2025-2026), none have passed.

Three compensating mechanisms fill the vacuum: (1) presidential executive orders, (2) state legislation, and (3) NIST soft law and technical frameworks. None are statutory-level instruments, but through common-law mechanisms such as “reasonable care” and “compliance presumption,” they harden into de facto requirements.

Executive orders (Trump-era, currently in force)

Section titled “Executive orders (Trump-era, currently in force)”
  • EO 14409 (2026-06-02) — classified cyber-capability benchmarking and a voluntary pathway for up to 30 days of federal pre-release model access
  • EO 14179 (2025-01-23) — “Removing Barriers to American Leadership in AI”; revokes EO 14110, launches 180-day Action Plan
  • Trump AI Action Plan + 2025-07 EOs — 25-page strategy document plus 3 companion EOs (Preventing Woke AI, Data Center, Exporting American AI)
  • EO 14365 (2025-12-11) — “Ensuring a National Policy Framework for AI”; launches preemption campaign against state AI laws (AI Litigation Task Force, BEAD funding conditions)

Executive orders (revoked, kept for historical reference)

Section titled “Executive orders (revoked, kept for historical reference)”
  • EO 14110 (2023-10-30) — Biden-signed “Safe, Secure, Trustworthy AI,” revoked by EO 14179 in 2025-01
  • NSPM-11 (2026-06-05) — replaces NSM-25 and directs accelerated, multi-vendor AI adoption, updated autonomous-weapons guidance, secure computing, assurance and TEVV across the defense and intelligence enterprise

Technical specifications (NIST, continuous across administrations)

Section titled “Technical specifications (NIST, continuous across administrations)”
DateEvent
2023-01-26NIST AI RMF 1.0 released
2023-10-30Biden EO 14110
2024-03 / 10Biden OMB M-24-10 / M-24-18
2024-07NIST Generative AI Profile (AI 600-1)
2025-01-23Trump EO 14179 revokes EO 14110
2025-04-03Trump OMB M-25-21 / M-25-22
2025-07-23Trump AI Action Plan + 3 EOs
2025-12-11Trump EO 14365 (state law preemption)
2026-06-02EO 14409 creates a voluntary frontier-model cybersecurity interface
2026-06-05NSPM-11 replaces NSM-25 and resets national-security AI adoption and assurance policy

May–June 2026 enforcement and implementation

Section titled “May–June 2026 enforcement and implementation”
  • On 21 May, the FTC announced a $930,000 settlement with Cox Media Group and two marketing firms over allegedly deceptive claims about an “Active Listening” AI-powered advertising service. The matter demonstrates continued use of FTC Act section 5 against AI marketing claims rather than a new AI-specific cause of action.
  • The FTC began enforcement of the TAKE IT DOWN Act on 19 May and sent warning letters on 20 May. The Act is technology-neutral but directly relevant to AI-generated non-consensual intimate imagery and platform removal workflows.
  • Colorado’s comprehensive AI Act reached its delayed effective date on 30 June 2026; Colorado also enacted sector-specific psychotherapy AI restrictions, signed 3 June and effective 12 August.

Primary enforcement source: FTC May 2026 releases.

Federal / state / local relationship (as of 2026-06-30)

Section titled “Federal / state / local relationship (as of 2026-06-30)”
  • Federal: deregulation and state-law preemption remain the baseline, while EO 14409 adds a voluntary national-security interface for cyber-capable frontier models and NSPM-11 establishes a defense/intelligence adoption architecture
  • State: 1,208 state AI bills introduced and 145 state laws enacted in 2025; the primary battlefield of AI governance
  • Local: TRAIGA explicitly preempts local ordinances; varies elsewhere

See subnational/us for details.

DimensionUnited StatesChinaEU
Comprehensive AI lawNone federal, state-law patchworkSectoral regulations + normative documentsHorizontal AI Act
Policy stabilityLow (EOs flip with administrations)High (central-party coordination)High (stable legislative process)
Enforcement bodiesNo dedicated regulator (FTC, EEOC, state AGs)CAC-ledMember-state MSAs + AI Office
Approach to frontier modelsEO 14409 voluntary federal interface + state laws (SB 53)Filing + TC260GPAI obligations + CoP
International postureCountering the Brussels EffectGlobal AI Governance InitiativeProjecting the Brussels Effect