AI regulation in United States (2026)

The White House transmitted legislative recommendations to Congress proposing a uniform federal AI framework covering child safety, community protections, and innovation promotion, explicitly aimed at displacing the growing patchwork of state AI laws. Represents the first formal White House attempt to establish a statutory federal AI governance structure.

President Trump signed an executive order directing the Attorney General to establish an AI Litigation Task Force to challenge state AI laws that conflict with federal policy and directing the Commerce Secretary to audit state AI regulations within 90 days. Escalates federal-state tension over AI governance by treating divergent state laws as an obstacle to national competitiveness.

Governor Newsom signed SB 53, requiring developers of large frontier AI models (trained with more than 10²⁶ FLOPS) to publish safety risk frameworks, report critical safety incidents to the state, and establish whistleblower protections for employees. Sets a de facto national compliance baseline given California's market size and signals ongoing state-level AI governance despite federal resistance.

The Trump administration published a comprehensive 90-plus-action federal strategy across three pillars — accelerating AI innovation, building AI infrastructure, and leading in international AI diplomacy and security. Formally replaces the Biden-era safety-first approach with a competitiveness-first national AI posture.

President Trump signed bipartisan legislation criminalizing the non-consensual publication of AI-generated intimate images (deepfakes) and requiring online platforms to remove flagged content within 48 hours of notice. Marks the first federal statute to directly impose criminal liability and platform takedown duties specifically for AI-generated content.

On his third day in office, President Trump revoked Biden's landmark AI safety executive order (EO 14110), eliminating mandatory safety-test reporting requirements for frontier AI developers and agency directives to appoint Chief AI Officers. Marked a sharp pivot from safety-focused federal AI governance to a deregulatory, innovation-first posture.

The Commerce Department's Bureau of Industry and Security published an interim final rule establishing the first export controls specifically targeting advanced AI model weights and high-performance computing chips, creating a global tiered licensing regime. The Trump administration rescinded the rule in May 2025 before it took effect, pledging a replacement with stronger controls.

NIST published NIST-AI-600-1, a companion to the AI RMF 1.0 that identifies unique risks of generative AI — including hallucination, data poisoning, and misuse — and proposes tailored risk management actions for each. Provides the primary voluntary US governance reference specifically for generative AI systems.

Colorado became the first state to broadly regulate AI systems that make 'consequential decisions' affecting consumers in employment, education, healthcare, housing, financial services, and legal contexts, imposing bias impact assessment duties on developers and deployers. Triggered a nationwide wave of similar state AI proposals and intensified debate over the need for federal preemption.

The most sweeping US government AI action to that point directed over 50 federal agencies to complete 100-plus tasks including mandatory safety-test result reporting for frontier AI models under the Defense Production Act, new NIST safety and red-teaming standards, AI-generated content watermarking guidance, and sector-specific risk reviews. Established the safety-first federal AI governance architecture subsequently dismantled by the Trump administration.

Amazon, Anthropic, Google, Inflection, Meta, Microsoft, and OpenAI agreed to voluntary commitments on internal and third-party safety testing, information sharing on AI risks, and watermarking of AI-generated content; eight additional companies joined in September 2023. Represented the administration's first major industry-facing AI governance mechanism and preceded the binding EO 14110.

After 18 months of development with input from over 240 organizations, NIST released the voluntary AI RMF 1.0 structured around four core functions — Govern, Map, Measure, and Manage — to help organizations identify and mitigate AI risks across the system lifecycle. Became the dominant voluntary US AI governance standard referenced in federal agency policy, state legislation, and international frameworks.

Enacted as Division E of the FY2021 National Defense Authorization Act, the National AI Initiative Act codified a coordinated federal AI research and development strategy, established the National AI Initiative Office within the White House, and created the National AI Advisory Committee (NAIAC). Gave the US government's AI coordination architecture its first statutory foundation.

The first US presidential executive order dedicated to AI directed federal agencies to prioritize AI R&D funding, develop technical standards, build an AI-ready workforce, and protect American AI advantages from strategic competitors. Established the deregulatory, competitiveness-first philosophy that has anchored US AI policy across both parties.