The Policy Landscape: Global AI Governance Divergence and Industry Influence

The Global AI Regulation Explosion

The AI policy landscape has undergone a dramatic transformation in just a few years. In 2016, there were virtually no AI-specific laws among G20 countries. By 2025, the picture was radically different. The United States had passed 25 AI-related bills—the most of any G20 nation—followed by South Korea with 17, and Japan, France, and Italy each with 9 to 10 laws. This acceleration reflects the urgent need for governments to address AI's economic, social, and security implications.

However, the expansion is uneven. While some nations aggressively legislate AI governance, others—such as Russia and Saudi Arabia—have passed very few or no AI-specific laws. This divergence signals fundamental differences in how nations approach AI regulation, ranging from precautionary frameworks (particularly in Europe) to more permissive, innovation-focused approaches (particularly in the United States and some Asian economies).

The Stanford AI Index Report 2026 emphasizes that policymaking is expanding but unevenly distributed. Similar to patterns in AI investment and research output, regulation concentrates in a small set of advanced economies. This creates a critical challenge: how can emerging economies establish effective AI governance when regulatory expertise and resources are limited?

The US Regulatory Paradox: Federal Retreat, State Advance

Perhaps nowhere is the governance divergence more apparent than in the United States. Federal AI policy shifted dramatically in 2025 toward deregulation, yet state legislatures moved in the opposite direction, passing a record number of AI-related bills.

Between 2020 and 2025, state-level AI bills surged from fewer than 10 to 150. California led dramatically with 62 bills over the full period—more than double any other state. Maryland (28), Virginia (25), and Utah (24) also demonstrated consistent activity. Notably, Missouri and Rhode Island had enacted zero AI legislation, revealing significant disparities in state regulatory capacity and priority.

State legislation covers diverse policy areas. California's recent laws exemplify the breadth: SB 243 regulates companion chatbots, AB 853 requires AI developers to include provenance data, and AB 621 extends existing deepfake protections. Utah's Mental Health Chatbot Act mandates disclosure of AI use and prohibits data sharing. Montana's Right to Compute Act takes a distinctly different approach, establishing a pro-innovation framework that protects computational rights. Texas's Responsible AI Governance Act focuses on high-impact uses, banning applications that incite harm or violate constitutional rights.

This diversity reveals a critical tension: how should AI be regulated? As a powerful technology requiring safety guardrails? Or as an innovation-driver deserving minimal regulatory friction? The absence of federal consensus has allowed states to experiment with different answers, creating what some call a "patchwork" and others call a "laboratory of democracy" for AI governance.

Congressional Attention and Industry Dominance

Congressional engagement with AI has grown exponentially. The number of witnesses testifying in AI-related hearings rose twentyfold from 2017 to 2025—from just 5 to 102 witnesses. This surge reflects AI's emergence as a central policy concern across national security, economic competitiveness, civil rights, and public safety.

However, witness composition reveals a troubling shift: industry now dominates congressional AI testimony. Industry's share nearly tripled from 13% in the 115th Congress to 37% in the 119th—making it the largest witness group. Over the same period, government witnesses declined from 35% to 10%, and academia fell from 26% to 15%. The "other" category (civil society and nonprofits) grew from 26% to 38%.

This composition change reflects both an opportunity and a concern. Industry representatives bring technical expertise and real-world implementation insights essential for informed policymaking. Yet the relative decline of government and academic voices raises questions about whether regulatory bodies have sufficient internal expertise to scrutinize industry claims and competing interests. The disproportionate industry presence also signals the sector's expanding ability to shape the regulatory environment in which it operates.

Congressional testimony focused on general AI governance (113 witnesses), national security and defense (74 witnesses), and finance and economic policy (36 witnesses from the House, 3 from the Senate). This concentration reflects pressing geopolitical concerns—particularly AI's role in national competitiveness and military capability—alongside growing awareness of economic and financial sector impacts.

Global Divergence: Precaution, Innovation, and Control

Global AI governance reveals three distinct philosophical approaches. Europe exemplifies precautionary governance. The EU AI Act, implemented in February 2025, established restrictive categories—banning predictive policing and emotion recognition systems—and requiring comprehensive risk assessments for general-purpose models. This "regulation-first" approach treats AI as inherently risky, requiring ex-ante controls.

The United States shifted toward deregulation in 2025. The Biden administration's Executive Order 14110 (2023), which anchored precautionary safeguards, was rescinded. The Trump administration's "Removing Barriers to American Leadership in AI" order explicitly oriented policy toward reducing regulatory constraints and promoting innovation. This represents a fundamental philosophical difference: AI should be treated as an innovation driver, not primarily as a risk vector.

China and other countries pursued strategic state capacity building. China's mandatory labeling rules for AI-generated content, coupled with domestic supercomputer investments and model development initiatives, reflect a state-directed approach focused on maintaining control over the AI ecosystem while enabling rapid development.

These divergent approaches create practical challenges for multinational enterprises. A system compliant with EU AI Act requirements may be over-regulated for US deployment but under-regulated for Chinese operations. Organizations must now navigate three distinct regulatory paradigms simultaneously, embedding flexibility and localization into their AI systems architecture.

Federal Agencies and Emerging Regulatory Consensus

Within the United States, federal regulatory activity has expanded significantly. The number of AI-related regulations grew from 1 in 2016 to 58 in 2025. However, this growth is concentrated among specific agencies: the Executive Office of the President (28 regulations in 2025 alone), the Commerce Department, and the Department of Energy.

This agency fragmentation reflects the cross-cutting nature of AI policy. The Department of Defense leads in contract spending (74% of total), driven by military and national security concerns. The Department of Health and Human Services and the National Science Foundation dominate grant funding, reflecting research and biomedical applications. This diversification across agencies suggests that AI governance is not consolidating into a single regulatory body but rather remains distributed, with each sector developing its own frameworks.

Key 2025 regulations illustrate this diversity. "Advancing United States Leadership in AI Infrastructure" directed federal agencies to expedite domestic data center development with clean energy requirements. "Preventing Access to Sensitive Personal Data by Countries of Concern" established cross-border data transaction limits. "Advancing AI Education for American Youth" mandated federal workforce development initiatives. Each represents sector-specific governance, not unified AI policy.

Implications for Organizations

The policy landscape demands operational flexibility. Organizations must simultaneously navigate EU precautionary requirements, US innovation-friendly frameworks, and emerging governance in other major economies. Rather than pursuing a single "global" AI strategy, enterprises should build modular AI systems capable of adapting to regional regulatory requirements.

Critically, organizations should also recognize industry's growing influence on the policy process. While industry participation in policymaking is natural and valuable, the relative absence of government expertise and academic voice raises questions about whether regulations will adequately address broader public interests. Responsible organizations should engage in policy advocacy not merely to reduce compliance burden, but to strengthen the regulatory foundation upon which sustainable AI deployment depends.

Featured image by A Chosen Soul on Unsplash.