Free AI hacking tool ARTEX is speeding up attacks on businesses
FBI warns free AI hacking tool ARTEX lets attackers reach admin access in under three hours
The FBI's warning of 13 November says attackers are using ARTEX, an open-source AI tool built to test networks for weak spots, against organisations worldwide. Barracuda has linked a cloud breach to the tool, and the FBI also cites a ConnectWise supply-chain breach that reached Microsoft 365 customer environments.
What to do this week
- Find every cloud key and password that may be exposed.Search code, shared drives and old configuration files, including earlier versions of files. Replace each one you find. Deleting a file does not make a leaked key safe, because copies may already exist.
- Revoke access you gave to outside companies.When a project with a vendor ends, remove the keys and accounts it was given. Ask every vendor that holds one of your keys where it is stored. Barracuda asked potentially affected customers to revoke and recreate their integration credentials.
- Require a second sign-in step on admin and remote access.Turn on multi-factor authentication, which asks for a code or app approval as well as the password, for cloud, email, admin and remote logins.
- Ask your IT provider to check internet-facing systems.Sign-in pages, password reset forms and APIs should need a login and limit repeated attempts. Configuration files and API documents that contain tokens should not be publicly reachable.
- Look for the signs in your logs.Watch for sudden bursts of requests, a cloud key used from an unfamiliar IP address, and sign-in attempts with made-up usernames. Keep logs long enough to show what happened. Barracuda said its logs were not detailed enough to rule out access, so it chose to assume the worst.
- Check whether anyone ran ARTEX from a Docker image.StepSecurity reports that copies pulled from the tomicdd/artex and congdonglinux/artex Docker images were backdoored. Remove any copy you cannot account for, and check the systems it ran on.
What happened
The FBI says attackers are using ARTEX, a free AI tool built for security testing, to break into business networks. In some attacks, they went from choosing a target to holding privileged access, the keys to an administrator account, in under three hours. Security firms use tools like this to attack a client's network on purpose, so the owner can find and fix the gaps. The attackers use the same kind of tool without that permission.
ARTEX is an open-source framework: its code is free for anyone to copy. It lets a set of AI agents run an intrusion test with little human direction. The original author later made the code closed-source, but copies, called forks, are still widely available, according to TechCrunch's report of the FBI warning. Unit 42, the Palo Alto Networks research team, counted 609 internet-facing ARTEX hosts as of October.
A related tool, Cairn, is also an open-source AI agent framework.
The full story2 more sections
How the attacks run
The tool does the slow, repetitive work that once needed a skilled person. It maps what a business has on the internet, such as sign-in pages, password reset forms and the interfaces software uses to talk to other software, called APIs. Unit 42 says attackers have reached cloud accounts through public password reset pages and exposed APIs, sometimes in less than a day. When a password or key has leaked into a public file, the tool can test it against other services within minutes. StepSecurity says the fastest way in is a credential someone left in the open.
Barracuda, a network and email security company, shows how this plays out. Its cloud environment was breached on 1 and 2 November, and the attacker reached about 10 cloud accounts over 44 hours. The way in was a cloud access key that Barracuda had given to an AI vendor for a past project. Barracuda believes the vendor's copy was taken in October from a configuration file in the vendor's GitHub repository. Barracuda found the intrusion on 4 November and admitted it had not retired the key: \"That is on us.\"
Who is affected
Any business with a website, a sign-in page, a cloud account or a service outsiders can reach is in scope. The FBI names financial services, government and technology companies among the targets.
ConnectWise, a US IT company, was breached in a supply-chain attack, which means the attackers used its access to reach its customers. The FBI confirmed earlier Microsoft findings that the attackers reached the Microsoft 365 cloud environments of at least 2,250 customers. ConnectWise says it cannot confirm that number. StepSecurity describes the same figure as sign-in attempts against more than 2,250 customer tenants, rather than successful access. Microsoft assesses the hackers are likely Chinese.
In South Korea, several financial institutions were hit, largely because of weak defences and no multi-factor sign-in. TechCrunch reports that the attacks stole the personal data of at least 68,000 people.
If your business relies on a supplier, that supplier's weak spots can expose your data too. Ask those suppliers in writing whether they have checked for the problems described here.
Timeline
-
HARD publishes its article, at High.
References9
- TechCrunch (). “FBI says Artex AI open source hacking tool used to breach networks in 'under three hours'”. FBI claims on timing, victims, the ConnectWise and South Korea figures, open-source copies, and Microsoft's attribution.
- Malwarebytes (). “Criminals are using ARTEX AI agents to speed up cyberattacks, FBI warns”. FLASH advisory dated 13 November, the three-hour wording, sector list, and the FBI's defender recommendations.
- BleepingComputer (). “Barracuda links cloud breach to ARTEX AI-powered cyberattack”. Barracuda's cloud breach: dates, 10 accounts over 44 hours, the vendor key, Barracuda's statement and customer steps.
- Unit 42 (Palo Alto Networks) (). “ARTEX and AI-Powered Red Teaming Tools: What Defenders Need to Know”. How attackers use ARTEX, timings, 609 internet-facing hosts as of October, and defensive recommendations.
- StepSecurity (). “ARTEX: Attackers Adopted This AI Pentesting Tool in Three Months. Leaked Secrets Are Its Fastest Way In.”. ARTEX appeared on GitHub on 26 July 2026, attacks from August, leaked secrets as the main entry point, the Docker image warning, and the 2,250 tenants figure.
- Unit 42 (Palo Alto Networks) (). “Attackers Use Autonomous AI to Target Semiconductor, Defense and Civil Society Sectors”. Cairn use against semiconductor, defence and civic targets. Read from the feed summary only; the full page was not read.
- Dark Reading (). “China-Nexus Actors Breach 10 Organizations in AI-Driven Attacks”. Cairn used in attacks on 10 organisations. Read from the feed summary only; the full page was not read.
- DataBreaches.net (). “FBI warns ARTEX AI hacking tool is powering global cyberattacks”. Feed summary only (full page blocked): FLASH advisory released on a Friday, and the ConnectWise supply-chain breach.
- Cloudflare (). “Adapting to AI penetration testing orchestrators: The Modern Security Model”. Defender countermeasures: static, non-expiring CI/CD tokens as a root cause, and rate limits and identity-gated access.
Situation Room articles are updated as events develop: updates are added to the timeline, and never edited away. Follow them by email or RSS.

