
The FBI issued a flash alert on Friday, November 13, about Artex, an open-source AI framework that turns a large language model into an automated intrusion tool. According to The Record, the FBI says Artex can “automate multiple phases of cyber intrusions at machine speed, from reconnaissance through data exfiltration,” with some intrusions going from initial targeting to privileged access in under three hours. The alert says the tool lets actors with minimal technical skill run complex, multi-target operations at once, and that it keeps spreading through forks, archived copies and a commercial version even though its GitHub repository was removed. We have covered earlier Artex incidents, including Microsoft’s ConnectWise attribution and the campaign against 200 organizations; this alert is the first we have seen from the FBI.
The Record reports that the FBI names financial services, government, technology, defense, semiconductor, energy, healthcare, manufacturing, telecommunications and legal organizations among the targets, and says Chinese models, especially the DeepSeek Flash family, are the most common in attacker-controlled instances. It also says Artex agents are good at finding weakly watched systems such as partner and supplier portals, mobile backends, employee applications and third-party integrations. We did not read the FBI alert itself, so these statements are as relayed by The Record.
Amazon published its own findings the same day. Its MadPot decoy sensors, set up to look like LLM inference APIs, received requests from Artex agents between October 3 and November 5, and those requests carried the agents’ own records of their campaigns. Amazon assesses with moderate confidence that the groups are disparate Chinese-nexus actors. It describes two campaigns: a large-scale theft of AI-model credentials, where stolen keys paid for the inference that powered the search for more keys, and an effort to get past identity checks at two cryptocurrency exchanges with real-time deepfakes. Amazon says it could not verify the outcomes the agents reported, and that the liveness-check vendor’s server-side verdict rejected the deepfake in every recorded attempt.
Counts differ between sources. The Record says the FBI cites Censys research showing over 1,600 publicly exposed Artex consoles. Censys’s own post says it saw 1,465 exposed consoles as of November 4, about 650 of them in two identically configured hosting blocks, and that real deployments are probably higher. The Record says the tool was released in August; Censys dates the GitHub release to July 26.
How to check if you’re affected
Affected products, per Amazon’s report, are internet-exposed AI applications and gateways that hold provider API keys. Amazon says the agents targeted LLM gateways and proxies (One API, New API, Veloera, Sub2API, LiteLLM) and AI application builders (Dify, FastGPT, Langflow, RAGFlow, Open WebUI, LobeChat, Ollama). Its playbook used default or unclaimed installs, anonymous management and provider-configuration endpoints, default administrator credentials such as the LiteLLM default master key, CVE-2025-3248 (Langflow), CVE-2026-28797 (RAGFlow), an authentication bypass through the isCliRequest header in One API-family systems, SSRF to cloud metadata services, and secrets left in client-side JavaScript, source maps, debug routes and environment files.
- Amazon says none of these techniques were new, and that the agents succeeded where basics were missing: default or hardcoded credentials, keys in code and public artifacts, unauthenticated interfaces and inference endpoints, and unpatched software.
- Its advice is to assume exposed secrets and misconfigurations will be found quickly, put AI applications, model gateways and inference endpoints under the same controls as other production systems, and watch for misuse of AI services through unexpected spend or unusual usage.
- Amazon lists indicator IP addresses for both campaigns (first seen October 14 to November 2) in an IOC table at the end of its post; check them against your logs.
- Amazon says it notified potentially affected AWS customers. If you run any of the products above on the internet, review the access and keys they hold; that step is our suggestion rather than a quote from the sources.
