A cybersecurity analyst monitors suspicious automated access attempts targeting a cloud-based system.
The FBI says attackers are using an open-source AI penetration-testing framework called ARTEX to run intrusions at what it calls machine speed. It also warns that the main danger is less-visible systems, such as partner portals, third-party integrations and app credentials. For Microsoft 365 and Entra administrators, one campaign tied to ARTEX makes this concrete.

A cybersecurity analyst monitors suspicious automated access attempts targeting a cloud-based system. What the FBI actually published​

The FBI released FLASH-20261113-001 on November 13, 2026. It is a TLP:CLEAR advisory titled "Threat Overview of ARTEX Autonomous Penetration Testing Framework." That matters for timing. ABC News reported on November 17 that the alert went out late the prior week. The FBI's own account on X says the advisory was released on Friday, 11/13, is not a new announcement, and has not been followed by any additional warning.

The FBI explained the timing as seasonal. As ABC quoted it, the bureau said that as the year ends, it wants Americans protected against criminals and state actors using AI to run familiar techniques at machine speed. The agency's own SI abbreviation appears in the ABC quote as written.

What ARTEX is​

ARTEX is an open-source, multi-agent penetration testing framework written in Go. It turns plain-language goals into commands that AI agents execute. In plain terms, an operator states an objective and a team of software agents works out the steps.

  • It has no built-in model. The framework does not ship with its own AI model. It connects to external ones.
  • Attackers favor Chinese-developed models. The FBI reports that attacker-controlled ARTEX instances most often use Chinese-developed models, led by the DeepSeek Flash family.
  • Guardrails are weak when the request is framed as testing. Industrial Cyber's summary of the advisory says the DeepSeek Flash family dominates because it balances capability with low cost and shows no observable guardrails when a request is framed as penetration testing.
  • The advisory cites no CVEs. SafeState's review of the advisory says it maps observed techniques to MITRE ATT&CK, lists indicators of compromise, and cites no CVEs. This is not a patch-Tuesday story. ARTEX is not a Windows or Microsoft vulnerability.

The "under three hours" claim, in context​

ABC quotes the alert saying ARTEX lowers the technical barrier to complex attacks, accelerates multi-target operations and exploits overlooked systems. The headline figure is that some intrusions progressed from initial targeting to privileged access in under three hours.

That figure has limits:

  • It describes some intrusions, not all of them.
  • It covers the path to privileged access, not necessarily data theft or full compromise.
  • SafeState's reading is that the advisory describes a change in tempo rather than a change in technique.

Defenders already know these techniques. The difference is that a team with a normal incident-response cycle may have less time to react.

Who is being targeted​

The FBI says malicious use has been seen since at least August. The sectors it lists are financial services, government, technology, defense, semiconductors, energy, healthcare, manufacturing, telecommunications and legal.

The advisory points to the parts of an organization that are least owned and least monitored. SafeState's summary lists partner portals, employee-facing applications, legacy services and third-party integrations. Anyone who has inherited a forgotten portal built by a business unit that no longer exists will recognize the problem.

Not every claim is backed by named victims. IBTimes notes that the advisory describes campaigns across multiple industries but does not identify a named hospital or government department confirmed to have been breached.

The Microsoft 365 angle: ConnectWise​

This is where the story touches Microsoft readers directly. The FBI Dallas office said ARTEX was used in the October ConnectWise supply-chain compromise, which exposed Microsoft 365 customer environments and involved attempted authentication against more than 2,250 downstream tenants.

Tom's Hardware reported on October 23 on Microsoft's findings. Its headline says China-aligned hackers used AI agents in the ConnectWise supply-chain attack. It says the agents attempted sign-ins at more than 2,250 Microsoft 365 tenants. It says some of the group's attacks reached privileged access in under three hours. Its subhead adds that automated password guessing failed, but that stolen app credentials needed no user sign-in at all. Only the headline and subhead of that article were captured, so the fuller technical detail is not independently confirmed here.

Two cautions apply:

  1. ARTEX is not a flaw in Microsoft products. The attackers abused third-party application credentials. The reporting I could verify does not show a vulnerability in Microsoft 365.
  2. Attempted sign-ins are not confirmed breaches. The 2,250 figure is the number of tenants where authentication was attempted. It is not a count of tenants that were compromised.

What admins should check​

The advisory's own recommendations were not available to me in full. The steps below are my own suggestions, based on the observed behavior reported above. They are not quoted FBI guidance.

  1. Inventory third-party app access. Review which vendor applications and service principals hold permissions in your Entra tenant, and remove any that are no longer needed.
  2. Look beyond interactive sign-ins. Review non-interactive and service-principal sign-in logs. Credentials that authenticate as an app will not appear in a user-login review.
  3. Watch for changes. Check for unexpected new app credentials, role assignments or group changes.
  4. Rotate credentials tied to compromised vendors. If a vendor you use has disclosed an incident, treat that vendor's app secrets as exposed until you have rotated them.
  5. Find forgotten internet-facing systems. Look for partner portals, legacy services and sales-support apps that no one actively monitors.
  6. Shorten your response time. Make sure logs are retained and alerts reach someone who can act on them quickly.

The tool will not go away​

The advisory material I reviewed does not say the original repository was taken down. The ARTEX developer details, the GitHub release in July, the October 8 removal and the commercial ARTEX PRO edition appear only in a research brief I could not verify against the FBI document. Treat those details as unconfirmed until the advisory text is checked.

What is verified is exposure. Censys counted more than 1,600 exposed ARTEX consoles in early November, and the FBI believes the true figure runs higher. Public scans cannot see deployments behind firewalls.

ARTEX and the week's other outages​

ABC reports that an FBI official said ARTEX does not appear connected to a run of recent incidents. These include a National Security Agency internal network problem, JPMorgan Chase banking outages, malware attacks via Optimizely, a nationwide ground stop and the Microsoft Outlook outage. Dallas News also reports that the tool has not been linked to recent outages.

This denial comes from ABC's reporting of an FBI official. It is not in the advisory itself. It is also a statement about what investigators currently believe, not proof that the incidents are unrelated.

Bottom line​

ARTEX shows that AI agents speed up known attack techniques and make them available to less skilled operators. The practical response for Windows and Microsoft 365 administrators is routine hygiene: tight control of third-party app credentials, full logging, and an accurate inventory of what is exposed to the internet. The advisory describes faster attacks, not a new kind of attack.

 

References

  1. Open-source AI tool has been used to automate cyber intrusions: FBI - ABC News - Breaking News, Latest News and Videos ABC News - Breaking News, Latest News and Videos 2026-11-17T20:35:00+00:00
  2. FBI Warns Hackers Are Using AI to Automate Attacks on Banks, Hospitals and Government Systems ibtimes.co.uk
  3. ARTEX AI Attack Framework: SafeBreach Coverage for FBI FLASH Alert FLASH-20261113-001 safebreach.com