ASAP
The Federal Administration Makes Legislative Recommendations for U.S. AI Policy, Leaving Questions Unanswered
The Trump administration has proposed “A National Policy Framework for Artificial Intelligence” (“the Proposal”). The Proposal, a non-binding set of legislative recommendations, encourages Congress to adopt legislation that emphasizes reliance on existing legal frameworks together with targeted interventions to protect existing rights while removing barriers to innovation.
As the accompanying press release lays out, the Proposal addresses the Trump White House’s six “most pressing” AI policy considerations:
- Protecting children by empowering parents to control their children’s digital environment;
- Safeguarding and strengthening American communities and small businesses;
- Respecting intellectual property rights by protecting American creators and publishers from AI outputs that infringe on their protected content while also supporting fair use;
- Preventing censorship and defending free speech and First Amendment protections by preventing AI systems from silencing or altering lawful political expression or dissent;
- Removing barriers to AI innovation and accelerating AI development and deployment; and
- Creating new jobs and expanding opportunities for American workers through AI training and education programs and workforce realignment studies.
The administration calls on Congress to adopt a federal policy preempting “cumbersome” state AI laws, stating that the Proposal can succeed only if there isn’t a state-level patchwork of regulations.
Is the Proposal consistent with prior AI-related efforts by the Trump administration?
The Proposal results from Section 8 of Executive Order 14365 issued in December 2025, in which the Trump administration directed the Special Advisor for AI and Crypto (David Sacks) and the Assistant to the President for Science and Technology (Michael Kratsios) to draft a legislative framework preempting “State AI laws that conflict with” the administration’s goal of “minimally burdensome” AI regulation. The EO was issued on the heels of a 99-1 Senate vote, in July 2025, to strike the 10-year “moratorium” on state-level AI regulations contained in the 2025 budget reconciliation bill (aka, the “One, Big, Beautiful Bill Act”).
When will there be federal AI laws aligned with the Proposal?
Although congressional action consistent with the Proposal is theoretically possible and attempts are expected, the 99-1 Senate vote indicates the overwhelming bipartisan disagreement with a state-preempting approach to AI regulation. The current political climate, the otherwise partisan divide in Congress, and upcoming midterm elections make it still more unlikely Congress will soon adopt AI laws consistent with the Proposal.
What does the Proposal imply about the Trump administration’s anti-discrimination policy?
The conspicuous absence of anti-discrimination protections in the Proposal is consistent with the Trump administration’s perspective on the validity of state-level efforts to combat bias in AI-based decisionmaking, and the implication that AI-based activity should not face scrutiny for bias.
EO 14365 declared that state AI laws “requir[e] entities to embed ideological bias within [AI] models.” The EO stated that banning algorithmic discrimination “may even force AI models to produce false results” simply to avoid disparate impact on protected groups, and therefore tasked a group of Cabinet members to “identify laws that require AI models to alter their truthful outputs.”
The Proposal takes this perspective one step further by detailing the steps needed to protect the six “American rights” enumerated above, and shielding other states’ rights—certain traditional police powers, zoning laws, heightened protections for minors and consumers and against fraud, and regulation of AI use in education, policing and other public-sector activity—from preemption. AI anti-discrimination laws, however, get neither a preemption pass nor a federal replacement. In contrast to the Tenth Amendment-based doctrine of concurrent authority, the Proposal calls for preemption of any “undue burdens” beyond the minimal regulations envisioned by the Proposal.
Significantly for employers, the Proposal simultaneously pushes back against state-level proposals of liability allocation between developers and deployers, requiring that states be prevented from “penaliz[ing] AI developers for a third party’s unlawful conduct involving their models.” In the discrimination context (and beyond), this could place a greater burden on employers using AI as part of their sourcing, hiring, and downstream employment operations, including the burden of explaining or justifying more-inscrutable AI outputs.
How should U.S. compliance strategies change based on the Proposal?
It is too early to determine how U.S. compliance strategies should be adapted based on the recommendations in the Proposal. However, U.S. employers can continue to develop compliance strategies based on the existing patchwork of state and local AI laws, study and comment on proposed AI legislation and regulation, and interpret generally applicable legal doctrines for AI use in the near term.
How will enforcement efforts change based on the Proposal?
The Proposal does not indicate whether or how enforcement efforts will change. While there has not yet been considerable enforcement activity from state agency officials or prosecutors, if immediate enforcement efforts were to be undertaken, they would likely be based on existing legal theories, such as anti-discrimination and data privacy.
What can U.S. employers do now?
In the absence of federal AI laws, U.S. employers can:
- Craft compliance strategies based on general principles (such as transparency in AI use, reasonable efforts to assess AI tools for legal compliance, training of AI users, and regular monitoring of AI-based outcomes) as well as requirements specific to the use of AI based on jurisdiction (such as reporting of impact assessments);
- Consider strategies to minimize the burden of compliance with a patchwork of laws, including through sandbox environments, pilots, and geo-gating; and
- Monitor proposed laws to prepare for significant changes and identify opportunities to submit comments when warranted.
These strategies can help balance the urgency for adoption with the risks associated with an uncertain legal landscape.
