Researchers warn LLM-driven auto-exploits could break patching as we know it

matthew_d_green · x · 2026-09-24

Security researcher attrc warns that LLMs hunting bugs 24/7 and auto-crafting weaponizable exploits break current patching practices — no org can keep up with weekly/daily patches. He cites a real incident against Chrome where open-source diffs exposed patched bugs, noting binary diffing and beta releases are equally vulnerable. Cryptographer Matthew Green amplified the warning.

Original post →

More from Safety

Safety channel →