Universal grammar isn’t “almost his entire body of work”—for the last thirty years he’s been at the forefront of the Minimalist program which approaches syntax from a very different direction. (And for twenty years prior to that he was working on X-bar theory, which was also a departure from universal grammar.) He’d still be considered one of the leading linguists of the past century on the strength of his non-UG work alone.

And Chomsky’s criticism of LLMs as a model of the brain’s internal language process is absolutely valid: while LLMs can imitate human languages, they can just as readily imitate unnatural constructed languages humans would never organically create. So they’re (at best) post-hoc predictive models, not explanatory ones—like Ptolemy’s epicycles, which could accurately predict the motions of the known planets but couldn’t generally distinguish between physically possible orbits and ones that would violate Newtonian mechanics, and thus provided no real insight into the natural world.