The White House released its National Policy Framework for Artificial Intelligence on March 20, 2026. Four pages. Seven pillars.
A promise of commonsense federal leadership on one of the most consequential technologies in human history.
Read it carefully and a different picture emerges.
The framework has been described as a blueprint for AI governance. It is not. A governance framework establishes rules, creates accountability, and builds enforcement mechanisms that give those rules teeth. This document does none of those things deliberately and by design. What it actually does is clear the legal field for the AI industry while calling that clarity governance.
Here is the tell. Among the nine major areas the framework leaves untouched are enforcement mechanisms and dedicated oversight. There is no new federal agency. No mandatory audits. No required testing for high-risk systems. No funding increase for the bodies that would theoretically enforce what little the framework does establish. The authors are not naive. They know that a standard without enforcement is a suggestion. They wrote it that way on purpose.
The liability section confirms it. The framework recommends that AI developers face no federal or state liability for unlawful third-party uses of their systems. Translated into plain language — if someone uses an AI tool to run a scam, commit fraud, or cause harm, the company that built and profited from that tool bears no legal responsibility. That is not a governance position. That is a legal moat built with congressional authority.
The preemption push tells the rest of the story. For the past year, states have been doing what the federal government refused to do — writing actual rules with actual teeth. Colorado’s AI Act. California’s transparency requirements. New York’s workforce protections. The framework targets these specifically, asking Congress to override them in the name of avoiding a patchwork of regulations. What they call a patchwork, the rest of us might call accountability trying to find a foothold wherever it can.
The carveouts are instructive. Children, fraud, consumer protection — these areas were preserved because the political cost of removing them would have been immediate and visible. Everything else, the algorithmic discrimination protections, the workforce displacement requirements, the environmental oversight, the high-risk classification systems — cleared. Not because those risks don’t exist. Because the industry finds them inconvenient.
One privacy organization summarized the entire document in six words. Protects AI companies, not people. That sentence deserved to be the headline of every analysis written about this framework. Instead it appeared as a footnote.
The American public deserves to understand what is actually being proposed here. Not the pillar headings. Not the innovation language. The architecture underneath it. A system where the technology moves at maximum speed, the profits flow to a concentrated set of actors, the risks are distributed across everyone else, and the legal structure is specifically designed to make accountability difficult to establish and nearly impossible to enforce.
That is not a framework for AI governance. That is a framework for AI without governance wearing the right clothes to get through the door.
The battle for who controls the rules of this technology is not coming. It is already underway. And right now the people writing the rules are the people who benefit most from having as few of them as possible.
Post Library – Intelligent People Assume Nothing
