I read the EU AI Act properly and I want to talk about what it actually does versus what people claim it does

I work in technology policy. Over the last year I have watched a lot of commentary about the EU AI Act that does not reflect what the legislation actually says. Technology companies claim it will kill innovation. Civil liberties groups claim it does not go far enough. Both positions are taken more confidently than the evidence warrants.

What I actually found when I read it: it is a risk-tiered framework that treats a medical diagnostic AI very differently from a content recommendation algorithm. The high-risk category requirements are significant but they apply to a narrower set of use cases than the headlines suggest. The prohibited uses are fairly specific and not particularly controversial.

I am not going to tell you it is perfect legislation. But I am frustrated by the quality of public debate about it. What specific provisions do people actually have questions about? I would rather have a real conversation about the tradeoffs than keep reading takes from people who have clearly not read past the executive summary.