Cairn AI agents found and exploited a zero-day, Amazon says

Saved work

Amazon Threat Intelligence says AI agents using Cairn discovered an unpatched flaw and used it to take control of a university server.

Amazon’s November 20, 2026 report describes operations against universities in East Asia outside mainland China during September 8–10. Investigators could read the attacker’s operational history because a publicly accessible Cairn instance lacked authentication.

The operator gave Cairn target addresses and asked the agents to gain shell access, or command-line control of a server. The agents selected and adapted their attack steps.

According to Amazon, they moved from mapping the application to writing files on the server in about 11 minutes. In a later session, they obtained root access, giving them administrator-level control.

Amazon links the operator to China with moderate confidence. Two other operations in its report failed to reach their objectives.

MONAI Label 0.8.6 contains the fix

MONAI Label is open-source software for annotating medical images and building AI annotation models. Cairn was the framework used to conduct the attack; the exploited application flaw was in MONAI Label.

The project’s security advisory describes a path traversal vulnerability in its upload handling. A caller could make an uploaded file land outside the intended storage directory, anywhere the server process had permission to write. The advisory warns that this can lead to code execution or disruption of the service.

Authentication is disabled by default, so a client that can reach the vulnerable server can exploit it in that configuration. Enabling authentication reduces exposure, but the advisory says an account with the annotator role is still sufficient to exploit the flaw.

The project’s release notes identify the vulnerability as CVE-2026-133968 and list versions through 0.8.5 as affected. MONAI Label 0.8.6 is a security release that includes the fix, and the package is available on PyPI. The release notes urge all server deployments to upgrade.

Keep MONAI Label on a trusted network after updating

MONAI Label’s maintainers warn that the application should run only on a closed, trusted local network. Their current security notice says other known, unpatched issues remain and explicitly warns against exposing either the server or a proxy in front of it to the public internet.

If your organization runs MONAI Label, confirm its installed version and network access rules. The maintainers also recommend running it without administrator privileges and limiting its file-write access to the data storage directory.

The maintainers recommend reviewing previously exposed servers for unexpected file changes before returning them to service. This applies to 0.8.5 and earlier deployments that were reachable from an untrusted network.

Businesses have exposure beyond medical imaging

In a separate September report, Gambit Security described a financially motivated campaign using Cairn alongside other AI tools to attack online retailers. Its investigators reported stolen payment data and damaging cleanup actions.

An exposed business application can be tested repeatedly without an operator choosing every next step. Security reviews need to include custom applications, research systems and forgotten test services, as well as the main website.

Ask your IT provider or security lead to check three things and report which systems need an update, tighter access or investigation.

  • External access. Identify services reachable from untrusted networks and remove public access where the application is intended for internal use.
  • Permissions. Confirm application processes have only the access they need. A file-upload service shouldn’t automatically have permission to change unrelated system files.
  • Response ownership. Name who investigates suspicious file changes or unexpected processes, who can contain an affected service and how that person is reached after hours.

Frequently asked questions

Was the vulnerability in Cairn or MONAI Label?

The exploited application flaw was in MONAI Label, a medical-image annotation tool. Cairn was the framework used to coordinate the AI agents carrying out the intrusion.

Does enabling MONAI Label authentication fix the flaw?

No. Authentication limits who can access the vulnerable functions, but the project’s advisory says an account with the annotator role can still exploit the flaw. Apply the security update and follow the project’s network and permissions restrictions.

Can an updated MONAI Label server be exposed to the public internet?

The maintainers say MONAI Label should run only on a closed, trusted local network. That restriction still applies after updating to 0.8.6. Their security notice warns about other unpatched issues and says not to expose the server or a proxy in front of it publicly.


Get new small business insights by email

Practical ideas and useful articles to help you make better business decisions.

HelperX Bot

Not sure what to read next?

I can suggest related Tech Help Canada articles based on the topic you’re reading now.

Tech Help Canada Staff researches, writes, and reviews practical content for business owners and professionals. Our coverage spans business, marketing, SEO, technology, and the tools and systems people use to grow and operate online. We focus on clear, useful information backed by research, hands-on experience, and editorial review. Learn more about our team and editorial standards. Need help with something? Contact Us

Leave a Comment

Tweet
Share
Share
Pin
WhatsApp
Reddit
Email