Skip to content
NewCybercrime Report 2026 · Eight Chapters · A Post Is Not An Incident → Explore The ChaptersCybercrime Report 2026 → Explore

MONAI Label Datastore File Write and RCE (CVE-2026-133968)

CVE-2026-133968 is a MONAI Label datastore path traversal flaw that permits file writes outside the intended storage directories.

Reviewed by
Brandefense Analyst Team
Published
Reading Time
3 min read
CVE ID
CVE-2026-133968
Vendor
Project MONAI
Affected Product
MONAI Label
Exploitation
Yes
Published
2026-11-18

The MONAI Label advisory says those writes can lead to remote code execution or service disruption. It lists monailabel versions through 0.8.5 as affected.Project MONAI published its advisory on 18 November 2026. Amazon Threat Intelligence reported on 20 November that an actor exploited the underlying flaw during a September campaign. The project has merged a code fix, but its advisory lists no patched package release.

What it allows

How Does CVE-2026-133968 Lead to File Write?

MONAI Label uses caller-supplied identifiers to build datastore destinations. The vendor says two upload paths did not verify that the resolved destination stayed inside the intended directory. A reachable client could therefore place content wherever the MONAI Label process had write permission.

The vendor says a write to sensitive application files can permit code execution, while overwriting configuration can disrupt service. Authentication is disabled by default, so network reachability is the key exposure condition in that configuration. If authentication is enabled, the vendor says the ANNOTATOR role is sufficient. For this exposure decision, see the vulnerability intelligence platform.

What Does the Evidence Say About Exploitation?

Amazon Threat Intelligence says autonomous agents found and exploited the flaw in a medical imaging application during a September 2026 campaign. Amazon reports that a later session gained a shell on one server. Its research does not publish the application name in the narrative, but links the Project MONAI advisory; that advisory credits Amazon with the original report.

Amazon's report documents exploitation in the described campaign. Use the vendor's access controls while awaiting a patched package release.

CISA KEV catalog version 2026.11.20 did not list the CVE. That catalogue status does not negate Amazon’s observed exploitation. Teams can use vulnerability prioritization to account for confirmed use, network exposure and the lack of a released patch.

Who is affected

Which MONAI Label Deployments Are Affected by CVE-2026-133968?

Project MONAI lists the monailabel Python package at version 0.8.5 and earlier as affected. The advisory states that authentication is disabled by default. It also says authenticated users with the ANNOTATOR role can reach the vulnerable behavior.

Inventory the version actually running on each MONAI Label server, then check network reachability and authentication settings. Review which accounts hold the ANNOTATOR role. A package manifest alone does not prove which code the active service loaded.

How Should Teams Contain and Fix the Flaw?

Until a fixed package release is available, follow the vendor’s workarounds: enable authentication, withhold ANNOTATOR from untrusted users and keep the server off untrusted networks. Run the service with minimal file permissions limited to its datastore. These controls reduce exposure but do not remove the flaw from an affected package.

The vendor says the code fix is merged and will appear in the next MONAI Label release. When a fixed release is published, deploy it through the patch management process. Verify the running package version and confirm from the release notes that it contains the datastore path-confinement fix.

As our investigation guidance, inspect for unexpected writes outside the datastore and review service restarts or new child processes around suspicious upload activity. Preserve relevant logs and compare file changes with approved deployments. Treat those observations as investigation leads, not proof of this specific exploit.

Timeline

  1. 2026-09-08

    Amazon Threat Intelligence says the observed campaign began around this date; its research describes exploitation of the medical imaging application flaw.

  2. 2026-11-18

    Project MONAI published the MONAI Label advisory, listing affected versions through 0.8.5 and no patched release.

  3. 2026-11-20

    Amazon published its investigation; NVD published the CVE record on the same date.

What to do

  1. Inventory exposed MONAI Label services

    Record the running package version, network reachability, authentication setting and accounts with the ANNOTATOR role.

  2. Apply vendor workarounds now

    Enable authentication, restrict ANNOTATOR, remove untrusted network access and limit service write permissions to the datastore.

  3. Review possible compromise

    Check for unexpected file writes outside the datastore and suspicious child processes. Preserve logs for investigation.

  4. Deploy and verify a fixed release when available

    Confirm the vendor publishes a package containing the merged fix. Upgrade affected services, then verify their active versions and release notes.

What Brandefense can do

Frequently Asked Questions

Is CVE-2026-133968 being exploited?

Amazon Threat Intelligence reports exploitation during a September 2026 campaign and describes shell access on one server. This is an observed campaign, not a count of all affected organizations.

Which MONAI Label versions are affected?

Project MONAI lists monailabel 0.8.5 and earlier. Network-reachable services with default disabled authentication are especially exposed; the ANNOTATOR role is sufficient when authentication is enabled.

Is a fixed MONAI Label package available?

The 18 November vendor advisory lists no patched package version. It says the code fix is merged and will be included in the next release.

What should operators do before a package release?

Follow the vendor’s workarounds: enable authentication, restrict ANNOTATOR, block untrusted network access and minimize the service account’s write permissions.

Sources

  1. Path traversal in datastore upload endpoints allows arbitrary file writeProject MONAI
  2. Chinese-nexus actor’s Cairn AI agents found a zero-day on their ownAmazon Threat Intelligence
  3. CVE-2026-133968NVD
  4. Known Exploited Vulnerabilities CatalogCISA
ShareLinkedInX
Brandefense logo

Stay Ahead of Digital Threats.

Take control of your digital security with an exclusive demo of our AI-driven threat intelligence platform.

Open The Full Demo Form →