If your US federal AI compliance framework is calibrated against the Biden-era regulatory posture and you expect the trajectory of recent years to continue, the January 2025 policy reversal makes the operating context fundamentally different. US federal AI-related regulations grew from 1 recorded action in 2016 to 58 in 2025. The pace has remained steady through 2025. But the direction reversed in January 2025: the Trump administration rescinded Biden's Executive Order 14110 (the framework for safe, secure, and trustworthy AI development) and replaced it with "Removing Barriers to American Leadership in Artificial Intelligence," reorienting federal policy toward reducing regulatory constraints and promoting innovation. The volume continues; the posture has flipped.

The volume trajectory deserves attention before the directional reversal. The annual regulation count: 1 in 2016, modest growth through 2020, sharp acceleration after 2022 (consistent with the post-ChatGPT policy attention surge), and steady high pace through 2025 at 58 actions. The federal regulatory pipeline is now producing AI-relevant rules at a pace that 2016-2020 frameworks didn't anticipate. The compliance load implied by 58 actions per year is materially different from the 1-action-per-year baseline of nine years ago.

The agency distribution shows where the activity concentrates. The Executive Office of the President has been most active, issuing AI-related regulatory actions every year since 2016 and putting out 28 in 2025 alone. The Commerce Department and the Industry and Security Bureau have become more active in recent years, consistent with growing attention to export controls and AI supply-chain policy. The Department of Energy, the Department of Education, and the Securities and Exchange Commission began issuing AI-related regulations in 2023 or later. The federal regulatory voice on AI is multi-agency and centred on the Executive Office.

The directional reversal in January 2025 is the contested observation. Two interpretations of the reversal are defensible from the data, and the planning implication depends on which holds.

Interpretation 1: the reversal is a sustained policy posture shift. The argument: the Trump administration's first 11 months produced multiple executive orders consistent with a deregulatory innovation framework (Removing Barriers, Advancing AI Infrastructure, Promoting Export of American AI Technology Stack, Ensuring National Policy Framework for AI). The pattern is consistent. The administration is signalling a 4-year posture of deregulation, innovation acceleration, and state-preemption. Under this interpretation, the 2026-2028 regulatory environment will continue at high volume but in a deregulatory direction. Compliance frameworks calibrated against Biden-era precautionary posture will be over-calibrated.

Interpretation 2: the reversal is a partial shift with persistent precautionary regulation in specific domains. The argument: while the executive-order framing has shifted, regulatory activity in specific domains (export controls, national security, child safety, financial regulation) has continued at high volume. The Commerce Department's export control activity, the Justice Department's data transaction rules, and the Treasury Department's financial AI guidance have continued or expanded. Under this interpretation, the regulatory environment in 2026-2028 will be mixed: deregulatory in some areas, persistently active in others. Compliance frameworks need to engage with the mix rather than read a single signal.

The data does not adjudicate cleanly between the two interpretations. The 2025 record shows both deregulatory executive orders and continued substantive regulation in specific domains. The recommended planning posture: build for both scenarios. Compliance frameworks should reduce burden in clearly deregulated domains (general AI governance) while maintaining or strengthening compliance in domains where activity has continued (export controls, national security, child safety, financial AI).

Three structural implications follow for organisations setting US federal AI compliance posture.

The first implication: the executive order pace is itself a planning variable. The Trump administration's first 11 months produced an unusually high number of AI-relevant executive orders. The 28 EOP regulatory actions in 2025 are not noise; they are policy formation in progress. Compliance frameworks need to track executive order content, not just enacted legislation. Many of the 2025 actions signal direction rather than binding compliance requirements, but the signal precedes the binding requirements in many cases.

The second implication: the federal-state interaction is now itself a federal regulatory variable. The December 2025 Executive Order on the National Policy Framework directs federal agencies to challenge state AI laws and tie federal funding to state compliance. The federal compliance environment now actively shapes the state compliance environment. Organisations operating across multiple US states need to engage with both layers as interrelated.

The third implication: international coordination has shifted. The Trump administration's posture diverges from EU AI Act enforcement, China's AI Action Plan, and the G7 cooperative framework. Organisations operating internationally face a more fragmented regulatory environment than the 2024 framework anticipated. The US export-control framework, in particular, is now both broader (covering more AI hardware and software) and more selectively applied (with country-specific carve-outs and partner-country provisions). The "Pax Silica Declaration" announced in December 2025, a US-led AI and technology supply-chain cooperation initiative, adds another layer to international coordination that didn't exist 24 months ago.

The contested question for compliance strategy: does the volume of 2025 executive activity represent a sustained 4-year operating environment or an early-administration burst that will normalise? The historical pattern for administrations is that the first year produces the most policy signal, with subsequent years producing fewer new initiatives but more implementation detail. If that pattern holds, 2026-2028 will be characterised by implementation of 2025 commitments rather than new directional shifts. If it doesn't hold, if the administration sustains the high pace of new initiatives, the regulatory environment continues to shift through the term.

For US-operating organisations setting AI compliance strategy through 2028, the planning anchor needs to engage with both the volume signal (58 federal actions in 2025 is the new operating baseline) and the directional signal (the posture has shifted from precaution to innovation). Plans that calibrate against either signal alone will mis-position the organisation. The plan that engages with both produces an operating model that absorbs the actual federal regulatory environment as it stands in 2026.

Sources

  • Primary: Stanford AI Index 2026, Chapter 8 (Policy and Governance) 8.4 — hai.stanford.edu/ai-index/2026
  • Federal regulation tracking: Stanford AI Index 2026 federal regulation tracker — AI-related federal regulations 2016–2025
  • Executive actions: January 2025 EO "Removing Barriers to American Leadership in Artificial Intelligence"; December 2025 EO "Ensuring a National Policy Framework for Artificial Intelligence"
  • Policy framework: America's AI Action Plan, July 2025; Pax Silica Declaration, December 2025