Trump scrapped an order requiring federal AI model vetting, deepening regulatory fragmentation as 33 states enact 100+ artificial intelligence laws in 2026.
President Trump signed and then quietly scrapped an executive order that would have created a federal vetting process for new artificial intelligence models before they reached the public. The reversal is striking not because it happened, but because the order existed at all: the administration had built its AI posture on deregulation, and a mandatory pre-release review ran directly against that grain.
Details on why the order was eliminated remain thin, and the administration has not offered a public explanation. What reporting does confirm, per Marketplace, is that the administration was previously committed to a hands-off approach to AI policy, making both the order's brief appearance and its removal notable pivot points in federal AI governance.
The state patchwork
States have not waited for Washington to decide what it wants. The National Conference of State Legislatures tallies 33 states that have collectively enacted more than 100 AI-related laws in 2026 alone, a figure documented by Marketplace. By any historical standard, that is a fast-moving regulatory wave.
For companies building AI products for national audiences, the practical consequence is a compliance puzzle with dozens of pieces. Disclosure requirements differ by state. Algorithmic audit rules, where they exist, vary in scope and timing. Sector-specific restrictions in health care, finance, and education add further layers. Large incumbents treat the resulting overhead as a cost of doing business. Startups often cannot.
Mapping the ecosystem
Into this environment arrives a new tool called Mapping AI, designed to help researchers, policymakers, and engaged citizens track who holds what positions across the artificial intelligence governance landscape. Marketplace describes its architecture as part professional network, part collaborative reference database, a way to see connections between organizations, public figures, and regulatory stances in one place.
The project is co-led by Anushree Chaudhuri, a Cambridge doctoral candidate studying large-scale energy infrastructure, and Sophia Wang, a research associate at Outliers Fund, a venture capital firm focused on emerging technologies. Their backgrounds diverge sharply, academic risk research on one side, early-stage investment on the other, and that contrast is deliberate. Governance decisions about artificial intelligence are being made by people who rarely share a room, and the tool is an attempt to at least put them on the same map.
"I do think a lot of people our age and just in general are feeling a huge lack of agency in understanding how to navigate something that's changing so quickly," Chaudhuri told Marketplace. Her team frames it as a civic resource as much as a research one.
What this means in practice
Federal deregulation of artificial intelligence is not passive neutrality. When Washington declines to set baseline standards, it does not eliminate governance; it hands the task to state legislatures that vary widely in technical expertise, lobbying access, and industry influence. The resulting patchwork historically concentrates market power: companies large enough to run compliance operations across dozens of jurisdictions absorb the cost; smaller competitors often cannot, which gradually tilts competition toward incumbents.
California's CCPA arrived in 2018, and the compliance burden it imposed shaped the market for years before any federal law materialized. AI regulation could follow a similar arc, except the legislation is arriving across 33 states simultaneously rather than sequentially. An artificial intelligence review process at the federal level, even a limited one, would have created a single point of engagement. Its absence does not prevent regulation; it multiplies it.
What comes next
Washington's retreat from AI model oversight does not end the regulatory question; it relocates it. State legislatures, federal agencies operating under existing statutory authority, and eventually the courts will now do work that a coherent national framework might have handled more efficiently. Whether that distributed process produces workable, consistent rules or a decade of jurisdictional friction is the central open question heading into the second half of 2026.
---
FAQ
What was the executive order Trump scrapped on AI?
The order would have established a federal review process requiring evaluation of new artificial intelligence models before public release. It was notable because the administration had previously taken a permissive, market-first approach to AI governance.
Why does it matter that 33 states passed AI laws?
Each state law can impose different requirements around disclosure, auditing, or sector-specific restrictions. Companies must satisfy all applicable rules simultaneously, which raises compliance costs and tends to disadvantage smaller firms relative to large incumbents.
What is the Mapping AI project?
Mapping AI catalogs the organizations, policymakers, and public figures shaping artificial intelligence governance. Co-led by a Cambridge researcher and a venture capital research associate, it functions as both a professional network map and a community-edited reference resource.
Does the US have any comprehensive federal AI law?
As of mid-2026, it does not. Governance is distributed across sector-specific agency rules, a growing patchwork of state statutes, and executive branch decisions, the most recent of which was the reversal of the pre-release review order covered here.