FBI flash alert warns the ARTEX agentic AI framework is automating intrusions; Amazon says Chinese-nexus groups used ARTEX agents to steal AI-model credentials and attempt deepfake KYC bypass
Amazon Threat Intelligence said on 13 November 2026 that from 3 October to 5 November 2026 its MadPot decoy sensors, set up to look like LLM inference APIs, captured requests from AI agents carrying the operating records of campaigns run on ARTEX, an open-source agentic AI penetration testing framework.
Amazon says the most prominent campaign was a large-scale theft of credentials for AI models in which stolen keys paid for the LLM inference powering further credential hunting; targets included LLM gateways and proxies such as One API, New API, Veloera, Sub2API and LiteLLM, and AI application builders such as Dify, FastGPT, Langflow, RAGFlow, Open WebUI, LobeChat and Ollama. A second campaign tried to defeat identity verification (KYC) at two cryptocurrency exchanges with real-time deepfake video and forged ID photos; Amazon says the vendor's server-side verdict rejected the deepfake in every attempt the agents recorded.
Amazon assesses with moderate confidence that ARTEX was used and that the actors are disparate Chinese-nexus threat groups.
It says it notified potentially affected AWS customers, published IOCs, and that none of the techniques were new: the agents succeeded where basic hygiene was missing, such as default or hardcoded credentials and keys left in code. (source: Amazon Threat Intelligence)
According to The Record, the FBI released a flash alert about ARTEX on 13 November saying the framework can "automate multiple phases of cyber intrusions at machine speed, from reconnaissance through data exfiltration", that some intrusions progressed "from initial targeting to privileged access in under three hours", and that despite removal of the original GitHub repository it "continues to proliferate through community forks, localized variations, archived copies, and a commercial version". (source: The Record)
- Product
- ARTEX agentic AI penetration testing framework (abused); targeted LLM gateways and AI app builders including One API, New API, Veloera, Sub2API, LiteLLM, Dify, FastGPT, Langflow, RAGFlow, Open WebUI, LobeChat, Ollama
- Versions
- n/a — threat activity
- Exploited in Australia?
- unknown
- Patch to
- Amazon advice: review your environment against the IOCs Amazon published; remove default or hardcoded credentials and keys left in code. CyberStack suggestion (our own): do not expose LLM gateway or AI app builder admin interfaces to the internet, and rotate provider API keys held by those apps if they were exposed.
