Half-measures guarantee full failure.
Too much of the federal government runs on technology that predates the iPhone. Veterans wait months for benefits processed on a system that was built in the 1960s and still hasn’t been replaced. The Small Business Administration’s Paycheck Protection Program portal crashed, leaving loan-seeking small businesses hanging, due in part to the underlying legacy systems.
The House has allocated $500 million to fix this embarrassment through commercial AI tools. It’s a rare moment of sanity: acknowledge that Silicon Valley builds better software than the General Services Administration, then buy what works instead of building what doesn’t.
But here’s the catch. That money will accomplish precisely nothing unless the Senate passes a version of the House’s ten-year moratorium on state AI regulation. Without it, Congress might as well light half a billion dollars on fire and call it a day.
The problem isn’t technical; it’s political. While Congress debates federal AI investment, state legislators have been running amok. They introduced over 1,000 AI bills this year alone, following 600+ last year (so far). Most won’t pass, but enough will to create a complex regulatory burden that risks making European regulation look relatively straightforward.
Here’s what “modernizing federal IT” would look like under this emerging regime: Imagine upgrading the benefits system so veterans actually receive timely payments. Sounds reasonable, right? Wrong. That system would now trigger analysis under New York’s proposed AI audit requirements and Texas’s automated decision-making disclosure mandates. Each state has different definitions, timelines, and penalties. The federal government would spend more on compliance lawyers than software engineers.
Or consider streamlining federal hiring. Each state where the government operates would demand impact assessments, bias audits, and algorithmic transparency reports. California’s Consumer Privacy Agency estimates that its AI rules could cost more than $800 million in the first year alone. Now multiply that figure by up to 50 states, each crafting their own special regulatory requirements, and the scale of the threat becomes clear.
The result is a smaller, weaker pool of vendors charging premium prices for the privilege of untangling regulatory spaghetti.
There’s a lesson to be learned from the CHIPS Act. This was supposed to revitalize American semiconductor manufacturing. Instead, Congress turned it into a progressive Christmas tree, decorated with requirements for childcare facilities, environmental justice assessments, and unachievable construction requirements. Every “worthy” cause got an ornament on the CHIPs tree.
And what did we get? Companies that could have been manufacturing chips are instead manufacturing compliance reports. Intel delays projects. Some companies simply took their projects to places that actually let them build. What should have been a strategic response to Chinese competition became an exercise in bureaucratic self-sabotage.
In this instance, it would be the states (if they are allowed to), rather than Congress, that would be making a mess of a project, in this case, that of the AI-based modernization of federal government functions. The House, to its credit, included both the investment and the regulatory relief needed to make it work. It will simply pause the proliferation of state AI-specific rules during the federal modernization period.
Some senators will undoubtedly worry about appearing “soft” on AI regulation by supporting the moratorium. This is exactly backward. The onslaught of state-level regulation is growing, but only a few bills have passed so far. The smart play is preventing regulatory chaos before it metastasizes, not cleaning up the mess afterward. Early intervention beats emergency surgery.
Others will claim we need more studies, more committees, more stakeholder input. Translation: They want to punt the decision to someone else while half a billion dollars circles the drain. This is how Washington works when it doesn’t work, replacing actual decisions with endless process. Will the Senate enable the largest federal IT modernization in decades, or let state-level regulatory theater turn it into the most expensive compliance exercise in history?
The choice is binary. Support both the AI investment and the regulatory moratorium, or watch another federal initiative collapse under the weight of its own contradictions.
Half-measures guarantee full failure.
Christopher Koopman is CEO at the Abundance Institute. Neil Chilson is the head of AI policy at the Abundance Institute.