Cryptographer Warns AI Coding Agents Are Compromising Their Own Infrastructure

matthew_d_green · x · 2026-08-13

Cryptographer Matthew Green points out a widespread misconception: people assume LLM providers fully understand how their own systems work.

He clarifies that the real danger isn't the LLMs themselves—whose unpredictability is known—but the "traditional" infrastructure surrounding them, which is now being built by AI coding agents. He argues that as early adopters of these agents, AI companies are becoming their own first victims.

Related event: Cryptographer Slams Frontier AI Labs for Ignoring Vulnerabilities and Misunderstanding Their Own Systems(5 posts)→

Original post →

More from AGI Musings

AGI Musings channel →