Google security researcher James Forshaw recently disclosed on social media that he had found a “ridiculously stupid” UAC bypass in Windows, and joked that the affected component looked like AI-written code. He is holding back for now on the affected component, exploitation details, and version range. This is not his first encounter with Microsoft—during the Administrator Protection preview stage, he found nine bypass methods in one go, all of which were fixed before general availability.

Microsoft’s Ledger: This Doesn’t Count as a “Vulnerability”

The problem comes down to definitions. UAC governs administrator privileges: everyday applications run with a restricted token first, and only sensitive operations such as installing software or modifying system files trigger a prompt for approval. Once it is bypassed, local malware can simply obtain a full administrator token, then freely modify the registry, disable protections, and install high-privilege backdoors. But Microsoft classifies UAC only as a “defense-in-depth feature,” not a security boundary that isolates privileges—under its security servicing criteria, a single bypass is by default not eligible for a security update, does not trigger a bug bounty, and usually does not even get a CVE ID. Forshaw’s remark that “Microsoft doesn’t need to consider UAC bypasses” lays bare exactly this set of rules.

Old Mechanism Retires, New Line of Defense Steps Up

Fortunately, the fix channel has not been closed. Although Microsoft does not publish advisories, it continues to fix issues reported by researchers, and has now fully replaced the traditional Admin Approval Mode with Administrator Protection, enough to block some entry-level bypasses. Now that the hole has been closed, Forshaw is expected to publish a blog post in due course, complete with a technical analysis and a proof-of-concept program—by then, it will also become clear whether the “AI code” jab has any hard evidence behind it.