UPD: It seems that the bill itself is not about the AI, but there are some undisclosed amendments to the bill that aren't yet public and that are said to target AI specifically. My bad. On the bright side - my wish to see a legal definition of an "AI" may still be granted :-)<p>--- Original comment below ---<p>Whem. Either I'm misunderstanding something (I've only quickly skimmed through the bill, can't say I understood any much), or the title is clickbait-ish, and this is:<p>1) Not about any AI at all (and here I was hoping to see a legal definition of an "AI", haha). The definition of "automated decision system" seem to cover just about anything from some fancy ML stuff to your basic fortune(6) program. Basically, if a machine rather than a human (or something else that is not a machine) had picked something, then it's an "automated decision system". At the very least, I have not found any sophistication requirements (but I could've missed it).<p>2) Only seem to apply when it's about use of personally identifying information. If I got it right, then if you track someone and a machine makes a "decision about the individual that could have a significant impact on them", then it applies. Otherwise, not so much. Again, I could be wrong, missing something or not really understanding what I've just read.<p>I haven't really understood how this applies to moderation. I suspect it's about when moderation system is building an user profile and feeding it to an automatic decision system, but that'd be weird (because every single mailhost with an anti-spam system would be affected).<p>Search engines, on the other hand? If I got it right, only those that do "personalized results" are affected.<p>So, some LLM processing search hits for a query and making a decision on what's good result and what's a spam is seemingly not covered. Tracking cookies seem to be the core component of this thing, not AI.