• Sources: primary, discussion, discussion
  • Summary: GTIG and Mandiant found an exposed staging server hosting exploit_agent, a multi-agent framework wired to frontier and local LLMs that compares code before and after patches, drafts exploits and executes them against targets without human review. The 2026-11-16 report also says threat actors now use open-source offensive agent frameworks such as ARTEX, CAIRN, CyberStrikeAI and PentAGI in live intrusions, and that monthly zero-day exploits rose from a 2025 average of 8 to 11 in 2026, spiking to 22 in August. GTIG says these frameworks trade stealth for speed and leave recognizable traces.
  • Why it matters: The window between a patch release and working exploits for internet-facing software is shrinking toward hours, so patch latency on exposed services is now the main risk to manage.
  • Background: Patch diffing compares a vendor's fixed code with the previous release to locate the flaw a patch closes, a technique long used to build n-day exploits by hand. Offensive agent frameworks chain an LLM planner with scanners and exploit tools so one operator can run many intrusions at once. This digest has tracked ARTEX through the FBI FLASH alert on 2026-11-15 and Amazon's decoy findings earlier today, and Unit 42's report on Cairn on 2026-11-14.