Skip to content
NewA calmer, faster ZeroHour

Amazon: ARTEX agents stole AI keys and tested deepfake KYC

Original title: ARTEX: How Cybercriminals Use AI Agents to Scale Their Attacks

Security Affairsby Pierluigi Paganini ()4 min read1
Part of a story covered by 3 sourcesAmazon links ARTEX agents to LLM key hunts and failed deepfake KYCMerged summary & timeline
AI summary · x-ai/grok-4.7Summaries by Vani AI

MadPot saw agent traffic from October 3 to November 5; Amazon has only moderate confidence goals were met.

Amazon Threat Intelligence says criminals ran offensive campaigns with ARTEX, an open-source agentic penetration-testing framework. Between October 3 and November 5, MadPot decoys posing as popular LLM APIs received requests that included the agents’ own campaign records. The largest effort sought AI-model credentials and reused stolen keys in the framework’s model gateway; a second campaign tried real-time deepfakes against KYC checks at two cryptocurrency exchanges. Amazon assessed ARTEX use and China links with moderate confidence, notified AWS customers it believed were affected, and published IPs seen from October 14 to November 2.

Unconfirmed: Amazon has moderate confidence on ARTEX use and China-linked groups, and could not confirm the agents achieved their goals.

Key facts
  • Planner, worker, verification and compression agents shared a knowledge graph.
  • Targets included One API, LiteLLM, Dify, Langflow, RAGFlow, Open WebUI and Ollama.
  • Agents looked for CVE-2025-3248 in Langflow and CVE-2026-28797 in RAGFlow.
  • A deepfake pipeline on one consumer GPU reached 11.5 frames per second.
  • Recorded deepfake selfie attempts were still rejected by the vendor’s server.
Security assessment · AI, from the article

Impact: Theft of AI and cloud credentials; attempted KYC deepfake bypass

ProductAffected
Langflow—
RAGFlow—
What to do
  1. 1.Check logs for the IP addresses Amazon published, observed October 14 to November 2.
  2. 2.Review exposed admin pages, default keys and multi-provider API gateways named in the report.
Analyze this story with Vani AIImpact, who is affected and what defenders should do, in a new chat

Vulnerabilities mentioned

All
CVEVulnerabilityCVSSEPSSSignalsPublished
CVE-2025-3248
Unauthenticated RCE in Langflow /api/v1/validate/code
9.8100%KEVRansomwarePoC ×4
KEV
CVE-2026-28797
RAGFlow: In versions 0.24.0 and prior, a Server-Side Template Injection (SSTI) vulnerability exists in RAGFlow's Agent workflow Text…
8.7<1%PoC

From the original article

4 min read at securityaffairs.com

Amazon spotted AI agents running on ARTEX steal AI credentials and test deepfake ID bypasses. Amazon’s threat intelligence team published findings on criminals running offensive campaigns with ARTEX, an open-source agentic AI penetration testing framework. Between 3 October and 5 November, Amazon’s MadPot network of decoy sensors caught requests from ARTEX agents.

Security Affairs

By Pierluigi Paganini · securityaffairs.com

Read the full article at Security Affairs

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.