Every security update changes some code. By comparing old and new versions, a researcher or an attacker can work out where the weakness was and how to use it against systems not yet updated.
It matters because AI is speeding it up. Google’s threat researchers reported on 16 November 2026 an AI tool that turned new patches into working attacks with no person involved.
