The new way of selling 'AI' seems to be to push it as a bugfinder. The latest example being waved around as of yesterday includes, as one of its non-embargoed examples, what is patched by this #OpenBSD patch.
https://ftp.openbsd.org/pub/OpenBSD/patches/7.8/common/025_sack.patch.sig
The problematic thing is that this sort of bugfixing hasn't changed the commentary in the code, which stated that p points to the last linked list entry at the point of the added null check and can never be null. But it actually can be, if there was a sole linked list entry that ended up being fully encompassed and thus deleted.
So this kind of 'AI' use is going to give us a lot more comments-do-not-match-code maintenance headaches down the road.
(Both #NetBSD and #FreeBSD factor this out into a separate tcp_sack.c and do the linked list handling slightly differently without a 'previous' pointer.)