Each autumn, Apple's software releases mark a quiet reckoning with the vulnerabilities accumulated over a year of human ingenuity and human error. This cycle, iOS 27 and macOS 27 arrive carrying fixes for hundreds of flaws — from memory corruption to kernel exploits — while also confronting a newer kind of uncertainty: the security risks that artificial intelligence introduces into systems it is simultaneously helping to protect. The most telling moment in this release is not what Apple shipped, but what it chose to hold back — an AI agent capable of changing passwords autonomously, shelved no