Elon Musk’s so-called Department of Government Efficiency (DOGE) operates on a core basis assumption that the United States should run like a startup. So far, this mainly means a chaotic shooting and a desire for steam regulations. But without excessive AI, no pitch deck in 2025 is complete, and Doge is no different.

Artificial intelligence itself shouldn't get a pitchfork. It has real uses and can create real efficiency. Bringing AI into workflows is not inherent, especially if you know and are able to manage around its limitations. However, it is not clear any nuances that Doge accepts. If you have a hammer, everything looks like a nail. If you have the best access to the most sensitive data in the country, everything looks like input.

Wherever Doge went, AI kept dragging it away. Given the opacity of the organization, it is unknown exactly how it is used and where it is. But two revelations this week show how widespread (possibly misleading) the desire for AI is.

In the Ministry of Housing and Urban Development, an undergraduate student at a university is tasked with using AI to find where HUD regulations may go beyond the strictest interpretation of basic laws. (Although the Supreme Court recently transferred this power to the judiciary, institutions have traditionally traditionally had broad interpretation powers.) This is actually a task for AI, which makes sense for AI, which can synthesize information from large files much faster than humans can. There are some risks to hallucinations—more specifically, to spit out a model of citations that don’t actually exist—but humans need to approve these suggestions anyway. On one level, this is actually what generative AI is now pretty good: doing tedious work in a systematic way.

However, some are harmful to ask AI models to help remove administrative status. (Besides the facts, your mileage will vary in this regard because you think low-income housing is a social benefit. It needs to be given a timely detailed information, which means you can not only use the referee, but also write a rulebook for them. It is also so eager to please that it will confidently form something rather than refuse to respond.

If nothing else, it is the shortest way to maximize the authority of major institutions, with the opportunity to spread nonsense.

At least this is an understandable use case. The same cannot be said for another AI effort related to Doge. As Wired reported on Friday, early Doge recruiters were once again looking for engineers, this time “designing benchmarks and deploying AI agents in real-time workflows for federal agencies.” His aim is to eliminate thousands of government positions, replace them with proxy AI, and “release” workers’ ostensibly “higher impact” responsibilities.