US Launches Dual-Track AI Governance Push Amid Fierce State Pushback

In June 2026, the United States federal government initiated its most comprehensive effort to date to govern artificial intelligence, launching a dual-track approach that pits federal supremacy against a growing patchwork of state laws. The landscape is defined by two major developments: a bipartisan congressional framework dubbed the Great American AI Act of 2026 (GAAIA), and a new executive order focusing on cybersecurity and voluntary model reviews. Together, these tracks represent a defining moment in how the U.S. intends to regulate the frontier of algorithmic power.[1][3]

The central tension in this emerging regulatory regime is the battle over preemption. For the past two years, state legislatures in California, Colorado, and New York have aggressively filled the federal void, passing stringent transparency and safety mandates for AI models. The new federal legislative proposals seek to consolidate this authority in Washington, sparking fierce resistance from state lawmakers who argue that federal baseline rules will gut localized protections against algorithmic discrimination and copyright infringement.[2][5]

Claim: The Great American AI Act would create the first binding federal governance regime for frontier AI developers, superseding state-level efforts. Evidence: Released as a 269-page discussion draft on June 4 by Representatives Jay Obernolte (R-CA) and Lori Trahan (D-MA), the GAAIA targets "large frontier developers." The legislation defines this cohort as companies generating over $500 million in annual revenue that train advanced frontier models, a threshold that captures industry leaders like OpenAI, Google, Anthropic, and Meta.[1]

Under the proposed act, these developers would be subject to mandatory transparency reports, risk assessments, and critical safety incident reporting. The bill also mandates whistleblower protections and establishes a new Center for AI Standards and Innovation within the Department of Commerce. Penalties for non-compliance are severe, with fines reaching up to $1 million per violation, signaling a shift toward rigorous corporate accountability.[1][4]

Evidence Strength: The most heavily scrutinized mechanism within the GAAIA is its preemption clause, which is explicitly detailed in the draft text. The bill proposes a three-year sunset period during which state laws "specifically regulating the development" of AI models would be preempted by federal statute. However, the text attempts a delicate carve-out, preserving state laws of "general applicability" and state authority over activities occurring at or after a model's deployment.[1]

Counter-Claim: State lawmakers argue the preemption clause will create a dangerous regulatory vacuum. Evidence: On June 16, a coalition of state legislators submitted a formal letter to Congress demanding the removal of the three-year moratorium. They assert that the tech industry will weaponize the broad preemption language in federal court to strike down state measures addressing AI training on copyrighted works, child abuse content, and consumer privacy. The evidence supporting the states' concern is strong, given the historical tendency of federal preemption to invalidate adjacent state-level consumer protections.[2]

Claim: The executive branch is pursuing a parallel, lighter-touch regulatory track focused on national security rather than mandatory commercial compliance. Evidence: On June 2, President Trump signed the executive order "Promoting Advanced Artificial Intelligence Innovation and Security." The order represents a strategic shift from the administration's previously hands-off approach, elevating the National Security Agency and the Treasury Department into central oversight roles.[3][7]

The executive order establishes a framework for the federal government to review advanced "covered frontier" AI models prior to their public release. Crucially, this engagement is strictly voluntary. The final text of the order explicitly states that it does not authorize mandatory licensing, permitting, or preclearance requirements. This language was reportedly hardened after internal debates, reducing the government's requested pre-release access window from 90 days to 30 days to avoid stifling innovation.[4][6][7]

The evidence indicates that the administration's primary concern is the weaponization of AI by malicious actors. The order directs federal agencies to strengthen cyber defenses within 30 days and establishes an AI cybersecurity clearinghouse to coordinate vulnerability scanning. Furthermore, it directs the Department of Justice to prioritize the enforcement of existing criminal statutes against AI-enabled cyberattacks, reflecting recent data showing massive spikes in AI-assisted hacking attempts.[3][6]

Uncertainty: A significant gap in the current evidence pack is the precise definition of which models fall under federal scrutiny. The executive order directs national security agencies to develop a classified benchmarking process to determine what constitutes a "covered frontier model" and to assess "advanced cyber capabilities." Because these definitions remain classified and pending, the practical scope of the executive order is currently impossible to quantify.[6]

The legislative and executive tracks present contrasting philosophies of governance. The GAAIA seeks to impose mandatory, market-wide standards enforced by the Department of Commerce, heavily influenced by the compliance frameworks pioneered in California and the European Union. In contrast, the executive order relies on voluntary industry cooperation and leverages the national security apparatus to address specific threat vectors, reflecting a strategy aimed at maintaining global AI dominance without burdensome regulation.[1][3][5]

These U.S. developments are occurring against a ticking clock in Europe. The European Union's AI Act, which entered into force in August 2024, is scheduled to implement its strictest obligations for "high-risk" AI systems by August 2026. U.S. lawmakers and the administration are acutely aware that failing to establish a domestic governance framework could result in the EU dictating de facto global standards for AI development and deployment.[5]

The immediate future of U.S. AI policy hinges on the legislative viability of the GAAIA. While the draft enjoys bipartisan authorship, passing comprehensive tech regulation remains historically difficult. If Congress stalls, the three-year preemption debate will become moot, leaving frontier developers to navigate the complex, fragmented landscape of state laws taking effect in 2026 and 2027, alongside the new voluntary federal security protocols.[2][5]