ON THE COMING INDUSTRIALIZATION OF EXPLOIT GENERATION WITH LLMS

Summary from the TLDR InfoSec newsletter

The author of this post discovered a novel zero-day in QuickJS using Opus 4.5 and GPT-5.2. They set up a testbed in which the agents were tasked with achieving different objectives across systems with varying protections. The author succeeded in generating over 40 distinct exploits, including complex exploit chains. Experiments such as this demonstrate that tasks like exploit generation are increasingly solvable by LLMs with enough tokens.