Trump’s National Policy Framework for Artificial Intelligence isn’t Moving Us Forward
The Impact of Inconsistency on Innovation
Ah, the framework. The promise ring of policy documents.
Not an engagement ring, not a wedding ring…something that is neither and that neither party really wants.
As a former technology policy maker, I will stick up for frameworks for just a moment. Sometimes there are issues that are so unfamiliar or esoteric that some level of strategic guidance is required. It is also a possibility that frameworks could be required if a single policy intent is to be communicated across a large and diverse organization that the component parts will then build upon with their own policy documents. While they can sometimes be a necessary evil, policy frameworks are perhaps the most impotent of all “action” that could be taken in a policy setting.
On the heels of the Trump Administration’s release of its National Cyber Strategy, the Administration dropped the “National Policy Framework for Artificial Intelligence” last week. The four-page deceptively titled document lays out seven policy priorities for AI. The framework on AI comes at a moment when the Trump Administration continues to struggle to find its footing in AI policy and legislation. Over the past year the Administration has tried a number of policy and legislative actions on AI that have been met with a variety of support and skepticism.
State regulation moratorium in the One Big Beautiful Bill Act (defeated 99-1 in the Senate)
An Executive Order threatening to take legal action against states that regulate AI (invoking 10th Amendment objections)
The Stargate investments ($500 billion promised to build AI infrastructure)
The Genesis Mission (without funding)
The AI Action Plan
Executive order on “woke AI”
Executive order on exporting the American AI stack
Executive order on accelerating federal permitting for data center infrastructure
This is a lot of action on a single policy area for just one year and it is increasingly difficult to make sense of the White House’s priorities and their strategic direction for such an important technology. The confusion is made worse because many of the above-mentioned orders, strategies, and frameworks do not align. Trump’s AI Framework is not the same as, nor an extension of, the AI Action Plan. The Stargate investment, much ballyhooed at the start of the Administration, has largely fallen from the headlines, and the calls for a massive national effort under the Genesis Mission flag have largely been for show as no funding was provided. If it feels confusing, it’s because it is.
But the AI industry and AI consumers alike do not want a promise ring. What they want is what that final wedding band provides, stability and consistency. Understanding the direction of national policy and legislation on any technology in any industry is what really enables innovation. This particular promise ring is going to exacerbate that confusion because it is the rarest of frameworks, the one that passes the buck.
The National Policy Framework for Artificial Intelligence is really the White House throwing the responsibility for AI action on Congress after a year of inconsistency and criticism of its positions. The framework seeks to clarify a few of the more contentious points while making it crystal clear that Congress needs to act. The White House is still seeking its white whale, a moratorium on state regulation of AI, but has recognized that its previous attempts to halt such action will not be effective.
This framework is no framework at all. It’s not meant for federal agencies to build their strategies and implementation plans upon it. It’s meant as one last communication and clarification of its priorities and a call on Congress to act. However you feel about the regulation of AI, it is likely that the White House will, with this framework, stop putting out policy actions on AI and leave it to Congress. Congressional action will be slow, if at all, meaning that the AI industry is right back where it started; an uncertain and inconsistent policy and legislative environment whose ambiguity is doing more harm than any individual regulation.
Calls for Legislation
The Trump Administration is hardly one known for delegating its priorities to Congress. A flurry of first week executive orders and unilateral decisions to engage in foreign combat operations offer more than enough evidence. For the last year, the Administration has tried multiple avenues to create the AI policies and legislation it wanted, and it has little to show for it. Love it or leave it, the Biden Administration’s AI policy actions were few and they were direct. That clarity has not been delivered by the Trump Administration. It is worth noting that Congress does not make policy, it makes laws. Providing legislative recommendations to Congress is not unusual, but it is unclear why the White House did not simply call this what it is, calls for Congressional actions.
Also telling is that a White House that has been very active in AI policy making is now shifting that burden to a Congress that it has largely ignored on its priorities. Time will tell, but currently it appears as if the White House wants to shift an issue, on which its positions are largely not popular, over to Congress. If Congress fails to act, the White House can point out their failure. If they succeed, it can point to this framework document and take credit. Win/win.
The Administration should be careful what it wishes for. If it fully delegates the future path of the AI regulatory environment, it may not get what it wanted. The 99-1 defeat of the state AI regulation moratorium during the One Big Beautiful Bill Act passage was a strong signal of what the people and their elected representatives want. The actions by state legislatures is not to spite Trump, it is in response to calls by the people using AI to put some guardrails around them and avoid the ongoing court battles over social media caused by a vacuum of legislation around the use of those platforms.
In a midterm year where Republicans are widely projected to lose at least the House, the White House’s timing is curious. Already, Trump’s election bill, the “SAVE America Act,” is caught in Congressional purgatory as the Senate “debates” a bill it knows it will never pass. Trump’s unbreakable sway over Congress is waning and that it would pick now to shift the responsibility of AI action to Congress is curious. Uncertainty and inconsistency continue to rule.
What’s in It
The four-page document includes seven policy priorities and mentions the word “Congress” 26 times. The White House’s press announcement states there are 6 key objectives, but the full document reveals 7. The 7th is a press for the Administration’s ban on state AI regulation (which did not make the much shorter press release).
Of all the tomes in the Trump AI corpus, this one is perhaps the best thought out and most grounded. There will always be room to debate what direction we want to take policy decisions, but this is a well-constructed document that could lead to legislation. The seven priorities are:
Protecting Children and Empowering Parents
Safeguarding and Strengthening American Communities
Respecting Intellectual Property Rights and Supporting Creators
Preventing Censorship and Protecting Free Speech
Enabling Innovation and Ensuring American AI Dominance
Educating Americans and Developing an AI-Ready Workforce
Establishing a Federal Policy Framework Preempting Cumbersome State AI Laws
Here are a few noteworthy sections from the document:
Under the IP heading,
Although the Administration believes that training of AI models on copyrighted material does not violate copyright laws, it acknowledges arguments to the contrary exist and therefore supports allowing the Courts to resolve this issue. Similarly, Congress should not take any actions that would impact the judiciary’s resolution of whether training on copyrighted material constitutes fair use.
This is a major piece of information coming out of the White House and shows it is moving off a major part of its platform. That it “acknowledges arguments to the contrary” is a major breakthrough for those advocating for privacy and copyright laws and it opens the door for Congress to take action for which it may have previously incurred wrath.
Under the Protecting Children heading,
Congress should require AI platforms and services likely to be accessed by minors to implement features that reduce the risks of sexual exploitation and self-harm to minors.
This is very clearly a call for AI safety and assurance, a previously dirty phrase from the Administration. The number of child suicides related to generative AI use is appalling and this is the first sign that we will no longer tolerate this issue. A clear victory for the AI safety space.
Under the Preventing Censorship heading,
Congress should prevent the United States government from coercing technology providers, including AI providers, to ban, compel, or alter content based on partisan or ideological agendas.
The ongoing row between the Pentagon and Anthropic ignited a debate over who controls AI. This objective calls for an end to the government’s influence in how AI model’s output information. Another open door for Congress.
Finally, under the section left out of the press release,
Preemption must ensure that State laws do not govern areas better suited to the Federal Government or act contrary to the United States’ national strategy to achieve global AI dominance.
“Better suited to the Federal Government.” Interesting sentence. This is the section of the recommendations that the Administration likely cares about the most, the wedding band it is hoping to get. It was also likely written almost verbatim by the Administration’s Silicon Valley donors. The Administration remains hard over on preventing states from regulating AI “because it is an inherently interstate phenomenon with key foreign policy and national security implications.” Yes, but it is used by tens of millions of non-foreign policy and national security users every day. The Administration may try to make the foreign policy and national security argument but the commercialization of the technology to regular users and how those users interact with the technology makes that argument all but moot. Cybersecurity as an industry also has foreign policy and national security implications, but the Administration isn’t making the same noise. The motivation here is not aligned with what users of the technology have been asking for and pushing that burden to Congress might result in something the Administration did not intend.
Some people feel good just to get a ring on their finger, promise or not. When it comes to AI, we need the kind of consistency that comes with wedding bands, legislation. The White House is calling for that legislation but wants it its way. Congress represents the people and its job is to pass laws that their constituents want. This framework does a good job of clarifying a few topics that were controversial and gives Congress an open door it has never had to act on AI. How Congress responds, if it responds, will shape the AI industry in a way no executive order could. The question is how Congress treats this door.