Trump’s AI Action Plan: Deregulation on paper, industrial policy in practice?
Photo Credit: Getty
Donald Trump’s follow-up executive orders to his 2020 artificial intelligence (AI) offerings and the new Winning the Race: America’s AI Action Plan signal a welcome rhetorical shift on tech policy. But many of the same structural pitfalls that plagued both his and Joe Biden’s approaches remain, which will make it harder for future reformers to disentangle the public and private sectors in AI – as well as nascent adjacent sectors such as smart cities and the expanding Internet of Things that emergent AI will increasingly rule.
The second Trump administration deserves credit for rejecting the Biden-era obsession with “woke AI.” Trump’s executive orders–particularly those aimed at purging federal procurement of ideological bias, revising the National Institute of Standards and Technology’s (NIST) Risk Management Framework to drop DEI language, and potentially promoting open-source development–are all steps in the right direction. These efforts are best exemplified in the administration’s ringing appeal to “Remove Red Tape and Onerous Regulation” that leads off their AI Action Plan. But beneath the surface, some components of these orders still reflect a top-down, government-first posture, just with new talking points. Government funding and regulatory incentives, rather than competitive free enterprise, threaten to dominate the sector’s future no matter who is president.
Instead of reducing government’s role in shaping the AI economy, some elements of these executive orders risk locking it in. One directive calls for the Department of Energy (DoE) to build out federal AI infrastructure at government-owned sites. Vastly more infrastructure is urgently needed, but the GOP was, until recently, determined to abolish the DoE. Another component of the plan would expand subsidies for AI exports via the Export–Import Bank and the US International Development Finance Corporation. The former is a classic industrial policy instrument; the latter is something of a shock given last week’s $9 billion rescission package that in part cracked down on related international escapades. The “Build American AI Infrastructure” and “Train a Skilled Workforce” rhetorical components echo the same central planning stance conservatives rightly reject elsewhere.
The proposed DHS-led AI cybersecurity and information-sharing initiative could potentially double down on the surveillance-adjacent architecture that concerned critics during the Biden years, and could again under some future administration if not the current one. Meanwhile, the revised NIST framework to in part “eliminate references to misinformation, Diversity, Equity, and Inclusion, and climate change,” remains a tool for standardizing AI priorities, both policy- and risk-wise, from Washington outward, rather than prioritizing hands-off, objective market experimentation driven by competitive realities.
The lesson from decades of federal regulatory misadventures, even those with the best of intentions, is clear: when government drives, it sets innovations barreling sideways on distortionary paths or even dead ends. What’s needed isn’t a federal action plan, but a national pause, on subsidies, centralized frameworks, federal deployment and public-private AI governance clubhouses, and new tasks for the very alphabet agencies once targeted for termination.
The alternatives to promote innovation in AI, as in all innovation, are cost-free: privatize, liberalize capital markets, relax predatory and anti-consumer antitrust activism, end distortionary government subsidies and loans, avoid expensive “safety” regulations that derail competitive disciplines and make us less safe, and continue to reduce over-regulation generally.
A superior set of executive orders would systematically place strict constitutional limits on federal AI use, ban AI-related subsidies and PPPs, and decentralize development to actual private actors (not “partners” and contractors). AI should evolve through open markets and competitive discipline, not through strategic guidance from OSTP or mandates from DHS. In other areas, Trump is reducing the inappropriate use of guidance documents, policy statements and other forms of regulatory dark matter. AI, one of the most promising fields imaginable, ought not be threatened by such ill-founded regulatory regimes we are otherwise trying to abolish at the get-go.
The Trump administration has an opportunity here: not merely to oppose woke AI, but to dismantle the regulatory scaffolding that enables such ideological as well as economic capture of industry in the first place. That means cutting government out of AI to the greatest extent possible, owning up to the folly and any such attempt, letting the engines of competitive free enterprise govern.
For more:
CEI statement on AI Action Plan
From this author:
“Artificial Intelligence Will Merely Kill Us, Not Take Our Jobs”
“Careful: Misbegotten Government-Business “Blueprints” Can Lobotomize Artificial Intelligence”