August 2026 · 5 min read
US Federal AI Governance
White House EO Reshapes Compliance

On December 11, 2025, President Trump signed the "Ensuring a National Policy Framework for Artificial Intelligence" Executive Order (EO), asserting broad federal authority over state AI laws. The EO establishes a "minimally burdensome" national AI governance framework, directly challenging the fragmented patchwork of state-level AI regulations that the Administration views as obstructing national AI innovation and competitiveness.
Key Definitions
2025 Federal AI Executive Order A December 2025 executive order signed by President Trump establishing a "minimally burdensome" federal AI governance framework. It asserts broad federal authority over state AI laws through preemption, advancing a "One Rule" approach to replace the fragmented patchwork of state regulations.
Federal Preemption The constitutional principle that federal law supersedes conflicting state laws. The EO uses preemption to centralize AI governance, allowing federal AI frameworks to supersede state reporting requirements while conditioning federal funding on state AI policy compliance.
Background
On December 11, 2025, President Trump signed the Ensuring a National Policy Framework for Artificial IntelligenceExecutive Order, a sweeping federal action designed to implement a "minimally burdensome" AI governance strategy and to significantly curtail the authority of individual states to regulate AI. The EO builds on prior initiatives — from the attempted 10-year state AI moratorium originally included in the Big Beautiful Bill to the Administration's AI Action Plan — consistently emphasizing a "build, baby, build" approach to AI innovation. In announcing the EO, President Trump stated, "You can't expect a company to get 50 Approvals every time they want to do something," underscoring the Administration's concern that expansive state-level requirements create a patchwork of 50 different regimes that complicate compliance and risk embedding ideological biases into AI models.
The EO specifically identifies Colorado's Consumer Protections for Artificial Intelligence Act — scheduled to take effect in June 2026 — as an example of a state law that prohibits "algorithmic discrimination." According to the EO, such laws risk requiring AI systems to "produce false results in order to avoid differential treatment or impact." The Administration calls for a single national standard that preempts conflicting state laws while safeguarding key priorities such as child protection, copyright integrity, and community safety. The EO frames this unified framework as essential to ensuring the United States remains competitive in the global AI race.
Key Points
The following are the core takeaways that enterprises should monitor closely:
1. DOJ AI Litigation Task Force
A central feature of the EO is the creation of a Department of Justice AI Litigation Task Force "whose sole responsibility" is to challenge state AI laws viewed as inconsistent with national policy. The Task Force will pursue litigation against state laws that, in the Attorney General's judgment, unconstitutionally regulate interstate commerce, raise preemption concerns, or otherwise conflict with federal law. Close coordination between the Task Force and White House policy advisors is mandated to monitor and respond to emerging state AI laws that warrant challenge.
2. FCC Federal AI Reporting Standards
Within 90 days of the Commerce Secretary's report, the Chairman of the FCC is directed to initiate a formal proceeding to evaluate the establishment of a federal AI reporting and disclosure transparency framework for AI models. If adopted, such a framework would supersede and preempt conflicting state reporting and disclosure requirements, further centralizing AI governance at the federal level. Notably, nearly two dozen state attorneys general filed a letter on December 19, 2025 urging the FCC not to issue preemptive AI regulations.
3. Commerce Department Review of State AI Laws
Within 90 days, the Secretary of Commerce must conduct a review of state AI laws deemed to conflict with national policy, identifying any laws that require AI models to "alter their truthful outputs" or compel disclosures that violate the First Amendment or other constitutional provisions. The Secretary may also identify states whose AI laws promote innovation consistent with national policy.
4. Funding Restrictions as Federal Leverage
Within 90 days, the Secretary of Commerce must issue a Policy Notice specifying conditions for state eligibility under the Broadband Equity Access and Deployment (BEAD) Program. States identified as having onerous AI laws will be ineligible for certain BEAD funds, to the maximum extent permitted by law. Executive departments and agencies must also review discretionary grant programs to determine whether funding can be conditioned on states refraining from enacting or enforcing conflicting AI laws.
5. 36 State Attorneys General Oppose Federal Preemption
In a November 25, 2025 letter, a coalition of 36 state attorneys general urged Congress to oppose proposals that would restrict states from enacting or enforcing laws addressing perceived risks associated with AI. This letter reflects broad resistance among states to federal initiatives that could constrain traditional areas of state regulation, including consumer protection, civil rights, and public safety. States are expected to frame legal challenges to the EO around core federalism principles, questioning whether executive action — absent clear congressional authorization — can lawfully constrain state legislative authority.
6. Enterprise Compliance: Navigating a Dual Regulatory Landscape
The EO introduces significant challenges for enterprises developing or deploying AI tools across multiple states and jurisdictions. State and local laws in Colorado, California, Illinois, and New York City may be subject to challenge, yet multinational employers remain subject to robust non-US requirements — such as the EU AI Act, China's algorithmic and generative AI rules, and Canada's forthcoming requirements. Critically, preemption of new AI laws does not eliminate legal exposure: existing federal and state civil rights statutes already prohibit discrimination regardless of whether an AI-specific law applies. The EO does not prevent employers from voluntarily implementing responsible AI governance practices, and organizations that invest in proactive risk management frameworks will be better positioned to navigate stakeholder expectations and mitigate litigation risk.
The December 2025 Executive Order marks a pivotal moment in US AI governance, signaling a transition from voluntary principles toward a contested federal-state regulatory landscape. While the Administration pursues a unified national framework, the coalition of 36 state attorneys general signals that states will vigorously defend their regulatory authority. Enterprises should prepare for a prolonged period of legal uncertainty by building adaptive compliance programs, monitoring both federal and state developments closely, and investing in voluntary AI governance frameworks — including AI system inventories, risk classification processes, and lifecycle controls — that remain prudent regardless of which regulatory regime ultimately prevails.
OOMeta AI
OOMeta's AI governance platform helps enterprises rapidly build AI system inventories, risk assessment processes, and compliance documentation systems — ensuring readiness and competitiveness in a fast-changing regulatory environment.
Schedule a DiagnosticReferences
FAQ
What is the "One Rule" approach of the 2025 White House AI Executive Order?+
The EO establishes that companies should not need 50 separate state approvals for every innovation. It advocates a single national AI governance framework to replace the fragmented patchwork of state-level regulations that the Administration views as complicating compliance and embedding ideological biases into AI models.
What are the core mechanisms the EO uses to implement federal AI governance?+
The EO creates a DOJ AI Litigation Task Force to challenge inconsistent state AI laws, directs the FCC to evaluate a federal AI reporting and disclosure framework, and uses federal funding leverage by conditioning BEAD program eligibility on state AI policy compliance.
What is the role of the DOJ AI Litigation Task Force?+
The Task Force's sole responsibility is to challenge state AI laws inconsistent with national policy. It pursues litigation against state laws that unconstitutionally regulate interstate commerce, raise preemption concerns, or conflict with federal law. The Commerce Secretary must identify onerous state AI laws within 90 days for the Task Force to address.
How are states responding to federal AI preemption efforts?+
In a November 25, 2025 letter, 36 state attorneys general urged Congress to oppose restrictions on state AI lawmaking. States argue federal intervention infringes on traditional state regulatory authority over consumer protection, civil rights, and public safety. States are expected to mount legal challenges around federalism principles.
How should enterprises navigate the new AI regulatory landscape?+
The EO is not deregulation but regulatory reshaping. Preemption of new AI laws does not eliminate exposure under existing federal and state civil rights statutes. Multinational employers must also comply with the EU AI Act and other international requirements. Enterprises should invest in AI system inventories, risk classification, and lifecycle controls.
相关文章
OpenAI 承认 Astra 思维链更难监控:审计证据必须从模型推理搬到动作边界
OpenAI 在 Astra 系统卡中首次承认:模型对自身思维链的控制力增强,链式思维监控的可信度下降,隐蔽作弊可能无法被发现。三天后首席科学家 Pachocki 撰文称没有任何实验室已解决对齐与监控。当被审计的实体能控制审计所读取的推理,审计就不再是独立证据。
知道坏了,不知道是谁干的:七成企业无法定位肇事 Agent
Kore.ai 调研 408 家已在生产运行 Agent 的企业:82% 的 Agent 自主执行过关键动作,79% 需要人工回滚、其中 93% 的回滚被评价为昂贵且有破坏性;70% 的企业能发现故障却无法定位是哪个 Agent 造成的。可观测性≠可归因,没有身份绑定的动作证据,遏制、回滚与问责都无从谈起。
Agent 将从 15 个暴增到 15 万个:90% 的企业却看不到它们在生产环境做了什么
Okta《AI Agents at Work 2026》报告:88% 企业已遭遇 Agent 安全事件,90% 却没有可靠手段治理 Agent 在生产环境的真实行为。财富 500 强的 Agent 数量预计从不足 15 个暴增到 2028 年的 15 万个。
“人在环内”正在变成幻觉:2026 年 Agent 人工监督该怎么做
2026年4月 MIT 评论直言“人在环内”监督已成幻觉——人类无法核验 Agent 内部推理。HITL 是阻塞闸门、HOTL 是事后监督,出路是把监督模型从 Agent 代码里抽出来,放到治理层按风险分级执行。