Adjacent to the special advisors of the White House Science and Technology Policy Director Michael Kratzos (left) and AI and Crypto David Sachs (right), President Donald Trump will sign an executive order at the White House AI Summit on Wednesday, July 23, 2025 at the White House AI Summit at Andrew W. Mellon Auditorium in Washington, DC.
In the final hours before Congress passed a massive budget settlement bill just before his July 4 holiday, the senator voted 99-1 to remove the proposed suspension on state law enforcement on artificial intelligence. But just two weeks later, the White House decided to do exactly what the Senate refused. They try to discourage state regulations of AI.
In the “AI Action Plan,” released on July 23, the Trump administration is trying to spread state restrictions on artificial intelligence again despite a “no” vote in the Senate. And like the suspension struck by the settlement package, this attempt could threaten serious harm, raise serious legal issues, and violate the Constitution.
What is an AI Action Plan? What do you say about state law?
The AI Action Plan is the latest attempt by the Trump administration to “achieve global domination in artificial intelligence.” The plan is built on three pillars, with the first one being to dismantle “nasty” federal and state regulations. The plan envisages attacking state regulations in two ways, through federal funding and the Federal Communications Commission (FCC).
First, the plan recommends that the Office of Management and Budget (OMB) and other federal agencies “consider the state’s AI-regulated environment” when making decisions regarding “AI-related” federal funding to the nation. The Action Plan directs agencies to “limit funds if the state’s AI regulatory regime could interfere with the effectiveness of its funds or awards.” The scope of “AI-related” funding could be extensive and could fund programs such as funding broadband, education, cybersecurity.
Second, the Action Plan recommends that the FCC “assess whether state AI regulations interfere with obligations and the ability of agencies to implement the authority under the Communications Act of 1934.” The FCC is an important institution for regulating certain technologies, but it is not a full-service technology regulator. Certain authority is granted over landlines, radio signals and cable television. As a result, directing state laws regulating artificial intelligence to be evaluated may seem misplaced – and as explained below, yes.
Replacing state laws opens the door for harmful AI
The nation has stepped up to deal with the way AI impacts all parts of our lives. The state is leading the way in which it passes laws regulating AI in state governments and key areas of life, such as housing, education, employment and credit. The state is also engaged in robust discussions on how to approach concerns about chatbots, generative AI and deepfakes.
Progress at the federal level is limited as states have been busy dealing with the harms of AI. Instead, some federal policymakers are pushing to ward off state efforts, including moratoriums on settlement packages, now AI action plans. Both the failed moratorium and the AI action plan create guardrails that replace the protections in the states they drive away, opening the doors of unregulated and harmful AI across our lives.
A wide sweep of plans is ripe for political abuse
The plan’s instructions have been wiped out and we should be wary of assuming that the Trump administration will apply it narrowly. For example, some commentators suggest that when there is little “AI-related” funding, state laws are only affected if they “block” the “effectiveness” of federally funded programs. As a result, they argue that the preemptive authority on the plan is extremely limited.
Certainly, that’s what the plan says, but the key terms here – “AI-related”, “distance”, “effectiveness” are adaptable and subject to weaponization.
For example, “AI-related” programs may be included in AI education, infrastructure or responses that rely on cybersecurity threats posed by AI. So you will reach funds to support K-12 school technology, rural broadband deployments, or enhanced cybersecurity in local governments.
Similarly, “intrusion” in federal programs could mean simply imposing or blunting implementations, including directly targeting these programs or imposing test or design requirements on AI developers. Given the emphasis on the action plan that AI development must be “uninterrupted,” the Trump administration could potentially choose a lower bar. In that case, state laws that “disrupt” AI may cover civil rights reporting requirements, environmental or zoning laws that affect data center sitting, or privacy laws that protect personal information. There is no similar definition of directive for the FCC to assess state laws that “obstruct” the control of communications laws.
Its widespread scope is exacerbated by the administration’s history of using federal funds as a hug against states and other agencies. This involves retrieving funds from New York for researchers in support of “DEI” from universities for student protests due to disagreements over immigration policy.
The potential scope of the preemptive approach to action plans has been wiped out, opening the door for harm to AI, where states are regulating and taking steps to attack the state because they are suitable for the political agenda.
Is it legal to replace these laws?
As always, the legality of an action plan depends largely on its implementation details. However, if an agency reads the action plan’s directives to the fullest, each of the two preemptive measures raises important legal issues.
First, new conditions placed on “AI-related” funds could exceed federal powers under the constitutional expenditure clause. The expenditure clause allows the federal government to condition funds provided to states and regions. However, that permissions are not unlimited. A “clear notice” is required to be in the terms state at the time the funding is provided, and the terms must be related to the purpose of the funding.
Action Plans may violate both “clear notification” and limits of purpose. There is no clear notice whether the state “will not be aware of the conditions or not see what is expected” or whether the federal government will impose “post-acceptance or “retroactive” conditions.” That’s right here – the action plan retroactively condition existing federal funds in the assessment of state regulations of AI. Similarly, the objective of the terms – “unobstructed” AI development – is unrelated to many programs that could potentially be dominated, such as K-12 funding, broadband deployments, and cybersecurity funds.
Second, the FCC directives far exceed traditional authority under the Communications Act. Over 91 years, the Communications Act was amended to expand the FCC’s authority over certain technologies. Wireless spectrum including cellular services, Wi-Fi, broadcast radio and television. Cable TV. Beyond these technologies, the FCC’s powers are extremely limited, and the courts have repeatedly pushed back the FCC’s efforts to regulate new technologies beyond jurisdiction, including the regulation of broadband internet services recently.
In limited circumstances, some courts support preemptively in certain information technology laws under “non-regulated federal policies.” However, in these examples, the underlying technology was “individual” from what the FCC could regulate under the Communications Act, such as the ability to make calls over the Internet. Other courts then stressed that they could not preempt state laws on technologies that the FCC has no authority to regulate, and questioned whether “federal non-regulated policies” could actually preempt state laws.
What’s next?
AI action plans are not legal documents, not self-execution. Instead, it must be carried out by executive orders and agency actions. This means that supporters raise concerns and the institution has the opportunity to revise courses. The FCC must engage the public in the public before embarking on rules, and OMB and other federal agencies should be transparent in implementing federal funding terms under the action plan.
Seriously, Congress should step up and take oversight. Several committees, including the House of Representatives Committee on Energy and Commerce, the Committee on Oversight, and the Senate Committee on Homeland Security and Government Affairs, must have jurisdiction over these issues and hold hearings. These hearings should feature testimonies from state officials who work to combat harm from AI, including Republican governors and attorney generals who opposed the moratorium on the settlement package. The committee can ensure that agencies remain within legal and constitutional boundaries and maintain their state authority to deal with the rapidly emerging harms of AI.
