Everything is about to “go dark”
Matthew Green argues that AI-assisted vulnerability finding could make well-maintained platforms hard to exploit remotely, removing the pressure-release valve that has so far let agencies buy phone exploits instead of demanding exceptional access. If that exploit supply dries up, governments may push vendors toward intentional backdoors—weakening the systems they claim to protect.
The HN discussion usefully challenges the optimistic premise: AI-generated code may expand attack surface as quickly as scanners close it, while platform-level bugs and human compromise remain difficult. Other reactions point to binary transparency, open source, self-hosting, and client-side scanning as parts of the defensive response.