We've known for a while that many if not most scientific journals are not great at peer reviews. A lot of junk gets through. Adding AI might be a good first step to catch major errors, but I suspect that what we really need is more rigorous requirements for publication.