You probably can't even index it, depending on how you interpret AI. Any vector based system, or probably even tf-idf, could qualify as machine learning and thus AI.
Yeah, it definitely prohibits you from applying latent semantic analysis, so you could probably violate the license by indexing it with Elasticsearch, but that's a less common thing to do with source code than opening it up in an IDE. TF/IDF seems like a borderline case to me.
Putting the source in Github indexes it, as well as probably any github competitor. Hell, if you're not careful even things like Mac's spotlight might index it. Any web search engine will also index it.
