ARTEX: attackers' AI agents spilled full campaign histories into Amazon's fake LLM APIs
colinmcnamara · x · 2026-11-22
Security researcher Colin McNamara's long read parses the FBI's Nov 13 FLASH alert on ARTEX, viewed from the LLM gateway layer between agents and models.
- What happened: From Oct 3 to Nov 5, Amazon's decoy sensors posed as popular chat-completions endpoints. Attackers' AI agents treated them as their brain, sending system prompts, tool history and shared memory — full operational records of campaigns. Amazon assesses with moderate confidence the agents ran the open-source ARTEX attack framework.
- Key insight: Every agent decision ships its entire operational history to the model, so single requests can contain a whole campaign — nobody broke in; the agents handed over their notes.
- Diffusion: ARTEX was picked up by criminals and at least one state-aligned group within weeks. The techniques are old; cost and speed changed.
- Defense: The gateway is both what attackers steal/run on and what catches them. Real case: an ARTEX operator had two routes to DeepSeek — a reseller and a local shim — so blocking one hostname leaves the other open. Full write-up with actionable gateway-trap advice on his blog.
Related event: FBI ARTEX Attack Report Sheds Light on LLM Gateway Risks(4 posts)→
More from Safety
- teortaxesTex cites Yudkowsky's Riemann passage to contest the 'RH-capable AI can kill you' threat model — teortaxesTex · 2026-11-22
- 'Imagine if Washington handed the internet to Gates in 1998': an AI regulation warning — Dan_Jeffries1 · 2026-11-22
- Can Zscaler Become the Default Control Plane for Enterprise AI Agent Access? — thedealdirector · 2026-11-22
- Solo Developer Builds Kavdren, a Verification Layer for AI Agent Responses — Few-Ad-1358 · 2026-11-22
- Anthropic's $1.5B book settlement unpaid; call grows to raise it to $20K per book — TuhinChakr · 2026-11-22
- AI Safety Researcher: Positive AI Futures Require Transparent US-China Collaboration — LuizaJarovsky · 2026-11-22