Problem is the way the data is made available is a dick for ai to analyse:

• ai can't directly log in and view the files because age verification blocks it.
• even then, if you were to upload all files by manually d/ling them, most ai have file limits so you can only use a subset.
• with 3m+ files, you'd probably need to build an ai specifically for all the drops. Even then of/when more data comes out, you will have to expand the source data of the ai
• the redactions can be overcome sometimes with ai for names, but accuracy is not great unless there's some pixel bleed in the censorship box.
• Even then it's tricky and needs manual fine tuning to get even 80% accuracy, and getting better (95% minimum for instance is my threshold for days quality) is nigh on impossible.