Security alert

The FBI's ARTEX warning, explained in plain English

Last updated: November 18, 2026

The short answer

On 13 November 2026 the FBI published a FLASH alert (FLASH-20261113-001) about ARTEX, a free AI tool that criminals are using to break into companies. You give it a goal in plain words, and AI "agents" do the hacking for it, from scouting a target to stealing data. The FBI says some attacks went from first look to administrator access in under three hours.

This page sums up what the alert says and lists the FBI's advice in plain English. It's based only on the FBI's own document. Omna is not mentioned in it and is not endorsed by the FBI.

What ARTEX is

  • An open-source tool released on GitHub in July 2026, described by its developer as a testing tool for authorised security checks.
  • It has no AI of its own. It plugs into outside AI models, including Anthropic Claude, OpenAI GPT, DeepSeek, GLM, Grok and Kimi. The FBI says attackers mostly use cheaper Chinese models, especially DeepSeek.
  • The developer took it off GitHub on 8 October 2026. But copies, local versions and a paid "ARTEX PRO" are still out there.
  • In early November, researchers found over 1,600 ARTEX control panels open on the internet, and the real number is likely much higher.

What it has already done

  • Attacks on South Korean financial firms took personal information on about 67,000 people.
  • A supply-chain attack through ConnectWise Cloud Backup led to attempted sign-ins at more than 2,250 customer accounts.
  • Targets span finance, government, tech, defence, chips, energy, healthcare, manufacturing, telecoms and legal firms around the world.
  • The FBI says it lets people with little skill run complex attacks on many targets at once.

How it gets in

ARTEX goes after the systems nobody watches closely, like partner and supplier portals, mobile app servers, staff apps, old services and tools built by other departments. The ways in the FBI lists include:

  • Leaked passwords and keys: admin passwords left inside website code, stolen VPN settings, leaked logins, and cloud keys left where they shouldn't be.
  • Weak admin passwords and login pages that can be bypassed.
  • Common website flaws such as SQL injection, command injection and missing login checks on web APIs.
  • Trusted partners: breaking into a smaller supplier first, then using its access to reach the bigger company.

Once inside, it looks for more passwords and keys to go further, then gathers data such as databases, mailboxes and cloud storage to steal.

The FBI's checklist, in plain English

  1. List everything you have online, including side systems like partner portals, mobile backends, staff apps and old services.
  2. Lock down admin tools. Cloud admin, deployment and backup tools should only be reachable from trusted networks.
  3. Give apps and accounts only the access they need, and watch for new passwords or permissions being added to apps.
  4. Put limits on busy pages and APIs. Add rate limits, and alert when someone looks up an unusual number of records.
  5. Don't show detailed error messages to people who aren't logged in.
  6. Hunt for hidden keys and passwords in settings files, environment variables, code, deployment pipelines and databases. Scan for them continuously, and change any that have leaked straight away.
  7. Keep the evidence (sign-in and audit logs) if something looks wrong, and don't delete things before they've been saved.

The full alert also lists the internet addresses and warning signs to look for. Your IT or security team should read the original FBI document.

Where Omna fits, and where it doesn't

Omna does not stop hackers or tools like ARTEX, and it isn't a replacement for the checklist above. It helps with one part of step 6: keys and passwords leaking through the AI tools your own team uses.

AI coding agents like Claude Code and Cursor read whole folders, including settings files, .env files, logs and terminal output, and send them to an AI company. Developers paste the same things into ChatGPT to fix errors. Every key that leaves this way is one more copy outside your control.

If you think you've been hit

The FBI asks you to report it to your local FBI field office or the FBI Internet Crime Complaint Center (IC3). Save your logs first.

Related: AI data loss prevention and our free AI acceptable use policy template.

Sources

Note: This page summarises the FBI's public alert as of November 18, 2026. Omna is not mentioned in or endorsed by the FBI. This is general information, not security or legal advice. For your own systems, follow the original alert and your security team.