In case you haven’t heard, the Trump administration’s MAHA political appointees were caught using an LLM (likely OpenAI) to write their MAHA report.

I guess we found the federal employees who can be replaced by AI!

Snark aside, this really says a lot–none of it good–about the people who are involved in what passes for policy development in the Trump Administration. There’s the sheer laziness of it all: they couldn’t even be bothered to check their work*. Yes, things need to happen quickly, but you still must confirm what you’re saying is passable. It also shows a lack of understanding of what LLMs actually do, which is… not smart or good.

The worst thing is, however, that they clearly have no idea how to write arguments supporting policy positions. While there are arguments to be had about whether LLMs should be used at all for writing government position papers (or anything else for that matter), what this shows is that they lack any understanding of how to write supporting arguments. When they misconstrue others’ work and cite non-existent work, that ultimately is a failure of writing. Good policy writing, even in the service of bad deeds, doesn’t make things up or lie.

But what they did was create a document that looks like policy analysis–and that achieves their desired ends–but isn’t actually the result of analysis. It’s just the result of a set of prompts that are designed to produce something that looks like the result of analysis and thinking.

This is the future these assholes want, and it will not be a good one if they’re able to realize it.

*It gets worse: these bozos replaced the bad citations with… more bad citations.