‘I’m a Government AI, and I’m Here to Help’

Artificial intelligence in government is here, and here to stay—the only question is whether it will be a relief or a disaster for we the people. Part one of the rollout was using and misusing AI in a scattershot effort to cut programs through the Department of Government Efficiency in early 2025, and part two was integrating AI into Pentagon systems soon afterward. Both efforts became messy, politicized affairs. But an even bigger test may come in part three: weaving AI into the regulation processes that shape the life of every business, as well as individuals like you.

The Donald Trump administration took a small step in that direction recently with the launch of America.gov, which aggregates data from about 29,000 federal websites to help users navigate procedures like changing their address across benefit programs. Will laying a large language model atop the expanse of government rules make it easier to deal with, or generate useful information for employees running the thing?

We’ve been taught that America’s regulatory system is organized in legal codes, like traffic signals and road signs. In reality, the codes are more like an overgrown jungle, incoherent to any real person and often contradictory—three billion words in total, according to law professor Daniel Ho.

Setting artificial intelligence loose into the red-tape thicket, experts think, will untangle the knots and make government work better. This is simplistic and largely misguided. While AI can certainly identify duplication and contradictions, it cannot replace the rigidities of red-tape operating structures that preempt common sense as a matter of law.