AI tools and news, explained in plain words
Agents

What is ARTEX? The AI hacking tool the FBI warned about (Nov 2026)

ARTEX is an AI hacking agent framework. The FBI says attacks took under 3 hours; Amazon says its agents stole AI API keys and tried crypto deepfakes.

Checked and updated 21 November 2026 at 15:25 UTC

Updates

  1. Added Amazon Threat Intelligence's own report (Nov 13): AI API key theft from exposed AI apps, deepfake attempts on crypto ID checks, attribution and Amazon's advice.
FBI seal, used by The Record for its ARTEX alert story
Image: The Record

Key points

  • ARTEX is an AI-powered penetration-testing framework that turns a large language model into a team of hacking agents. It first appeared publicly in mid-2026 as open source.
  • In a FLASH alert released on Friday, November 13, 2026, the FBI warned that criminals and state-backed groups use it against banks, governments, tech, defense, healthcare and law firms, sometimes reaching admin access in under three hours.
  • Amazon says ARTEX agents ran from Oct 3 to Nov 5 stealing AI API keys from exposed apps like LiteLLM, Ollama and Open WebUI, and tried deepfakes on two crypto exchanges' ID checks. They succeeded mostly where basics like passwords and patches were missing.

ARTEX is an AI hacking framework: it connects a large language model to security tools and lets AI agents run the steps of a network break-in by themselves. On Friday, November 13, 2026 the FBI issued a FLASH alert warning that criminals and state-backed hackers are using it, and that some attacks went from first contact to admin-level access in under three hours.

The name has been all over the news since, so here is what ARTEX is, who has been hit, and what it means for regular people.

What ARTEX is

ARTEX started as an open-source "AI penetration testing" platform. Pen-testing tools are normally used by security teams to probe their own defenses. ARTEX adds AI agents on top: you give it a high-level goal and it turns that into machine commands, working through the stages of an intrusion at machine speed.

It works with almost any AI model. Reports name Anthropic's Claude, OpenAI's GPT, DeepSeek, GLM and Kimi. According to The Record, the FBI says Chinese models are the most common in attacks, with DeepSeek's Flash models used most.

The FBI's main worry is who can now do this. Its alert says the framework lets people "with minimal technical expertise" run complex attacks on several targets at the same time.

Who made it and is it still available?

The Record reports that a GitHub account called "Autumn-27" released ARTEX, and links a Wall Street Journal report naming a Chinese security engineer as the developer. On October 8 the developer posted that the tool had been abused, denied responsibility and later removed the repository. TechCrunch says the software went closed-source after the attacks. The FBI says the old open-source version is still "widely accessible" through forks, archived copies and a commercial version.

What attacks has it been linked to?

Target What was reported
South Korean financial firms Personal data of at least 68,000 people stolen; weak defenses and no multi-factor authentication
ConnectWise (IT software) Microsoft blames a likely Chinese group; the FBI says Microsoft 365 data of at least 2,250 customers was accessed. ConnectWise can't confirm that number
Barracuda (security company) Criminals used a leftover credential from a former partner; Barracuda reset all customer credentials
Japanese financial firms Several attacks, per The Record

Sectors named in the alert: financial services, government, technology, defense, semiconductors, energy, healthcare, manufacturing, telecoms and law firms. Amazon's security chief told The Record that such frameworks now run "multi-objective offensive operations" with little human help.

What Amazon saw inside ARTEX campaigns

Amazon published its own findings on November 13, in a post by its security chief CJ Moses. Amazon runs a network of decoy servers called MadPot. Some of them pretend to be popular AI model APIs. From October 3 to November 5, 2026, ARTEX agents sent requests to those decoys as if they were real AI services, and because AI agents carry their history with them, the requests included the agents' own notes about their campaigns.

What those notes showed:

  • Stealing AI keys to pay for more attacks. The biggest campaign hunted for API keys to AI models, then fed the stolen keys back into ARTEX so the victims paid for the AI doing the next round of hunting. Targets were internet-facing AI tools that store provider keys, including LiteLLM, One API, New API, Dify, Langflow, Open WebUI, LobeChat and Ollama. The agents also combed public code on GitHub, npm and PyPI packages, container images and paste sites for leaked keys.
  • Deepfakes against ID checks. A second campaign tried to pass the selfie and ID checks at two cryptocurrency exchanges with real-time deepfake video and edited ID photos. Amazon says the verification vendor's server rejected the deepfake in every attempt the agents recorded.
  • Who was behind it. Amazon says, with moderate confidence, that several separate Chinese-nexus groups ran these campaigns, based on Simplified Chinese prompts, China time-zone settings and their choice of targets.
  • What Amazon cannot confirm. It did not watch the attacks happen on real victims' systems and cannot verify the results the agents reported. It notified AWS customers who looked affected.

Amazon's conclusion is blunt: none of the techniques were new. AI agents just make attackers faster and more persistent, and they won where basics were missing, such as default passwords, keys left in code, AI servers with no login, and unpatched software.

Does it affect you?

The alert is aimed at organizations, not home users. You are mostly affected indirectly, when a bank, hospital or service that holds your data is breached. The good news is that ARTEX attacks rely on old weaknesses, according to TechCrunch: exposed or stolen passwords, weak protections and missing multi-factor authentication. The FBI also notes the attacks are "noisy", so they can trip alarms, but move too fast for slow responses.

What helps, for people and small businesses alike:

  1. Turn on two-factor authentication everywhere, especially email and banking.
  2. Use a different password for every account, ideally from a password manager.
  3. Remove access for old apps, former contractors and unused accounts.
  4. Watch for breach notices from companies you use, and change passwords when you get one.
  5. If you run your own AI tools, such as Ollama or Open WebUI on a home server, keep them off the open internet or behind a login, update them, and set a spending limit on any AI API keys so a stolen key can't run up a big bill.

ARTEX is part of a bigger shift toward AI agents in security, on both sides. Google's own research on this is in our Google AI hacking agents report explainer, OpenAI's security-focused model is explained in GPT-6 Cyber, and if the term is new, start with what is an AI agent.

AI Tools This Week is not affiliated with the FBI or any company named here.

Frequently asked questions

What is ARTEX AI?

ARTEX is software that connects an AI model, such as Claude, GPT, DeepSeek, GLM or Kimi, to hacking tools and lets AI agents carry out the steps of a network break-in on their own. It was published as a security testing tool, but the FBI says attackers now use it.

Is ARTEX a threat to my home computer?

The FBI alert is about organizations, such as banks, government agencies and IT providers. Regular people are mostly affected when a company that holds their data is breached. Using unique passwords and turning on two-factor authentication still protects you most.

What did Amazon find about ARTEX?

In a November 13, 2026 post, Amazon said its MadPot decoy sensors caught ARTEX agents from October 3 to November 5 hunting for AI API keys in internet-exposed AI apps and trying to beat crypto exchange identity checks with deepfakes. Amazon links the activity, with moderate confidence, to several Chinese-nexus groups.

Is ARTEX still on GitHub?

No. The Record reports the developer posted a statement on October 8 saying the tool had been abused and later removed the repository. The FBI says forks, archived copies and a commercial version are still in circulation.

Sources

  1. Amazon Threat Intelligence observes threat actors running cybercrime campaigns with ARTEX AI agents (AWS Security Blog, CJ Moses, Nov 13, 2026)
  2. FBI warns of AI tool 'Artex' automating cyberattacks (The Record, Nov 13, 2026)
  3. FBI says Artex AI open source hacking tool used to breach networks in 'under three hours' (TechCrunch, Zack Whittaker, Nov 16, 2026)

The AI Tools This Week Desk writes and edits every article, checks claims against primary sources and updates stories when the facts change.

Related

Mentioned in