CVE-2026-100846: MONAI: unsafe pickle deserialization allows RCE
HIGHPoC AVAILABLECISA: ATTENDMONAI versions before 1.5.2 contain an unsafe deserialization flaw in algo_from_pickle (monai/auto3dseg/utils.py), which passes the contents of a .pkl file straight to pickle.loads, so a crafted file executes attacker code with the application's privileges. The package has 126 downstream dependents and 14 other tracked CVEs, and it is widely used in medical imaging pipelines where hosts often hold sensitive patient data. Exploitation likelihood is currently low: EPSS is 0.00292 (top 80th percentile), it is not in CISA KEV, and no public exploit or Nuclei template exists. Still, pickle RCE is trivial to weaponize once a victim loads an attacker-supplied file, and the CVSS of 8.8 reflects full confidentiality, integrity and availability loss. Upgrade to MONAI 1.5.2 or later, and until then block untrusted .pkl files from reaching Auto3DSeg workflows.
What is the risk?
High severity (CVSS 8.8) but low near-term exploitation likelihood (EPSS 0.00292, no public exploit, not in KEV, no scanner template). Exploitation requires the application to call algo_from_pickle on an attacker-controlled file, which means user interaction or an exposed file-ingestion path. Once triggered, impact is arbitrary code execution with full CIA impact. Exposure is concentrated in research and clinical ML environments that exchange Auto3DSeg algorithm artifacts, which are often shared between teams and downloaded from external sources. The OpenSSF Scorecard of 7.9/10 and package risk score of 26/100 indicate a reasonably maintained project, and the repository is actively pushed to.
How does the attack unfold?
What systems are affected?
| Package | Ecosystem | Vulnerable Range | Patched |
|---|---|---|---|
| MONAI | pip | — | No patch |
Do you use MONAI? You're affected.
How severe is it?
What is the attack surface?
What should I do?
1 step1) Upgrade MONAI to 1.5.2 or later (pip install -U monai) and rebuild container images. 2) Until patched, do not call algo_from_pickle on files from untrusted or unverified sources; restrict write access to Auto3DSeg working directories and algorithm template stores. 3) Verify integrity of shared algorithm artifacts with checksums or signatures and keep them in access-controlled storage. 4) Scan pickle artifacts with picklescan or fickling before loading, and prefer non-executable formats where possible. 5) Run training and experimentation jobs with least privilege in sandboxed containers with restricted egress and no ambient cloud credentials. 6) Detect: hunt for unexpected child processes (shell, curl, python -c) spawned by MONAI/Auto3DSeg Python processes, unusual outbound connections from training hosts, and unexpected .pkl files in project directories. 7) Inventory MONAI usage via SBOM and dependency scanning.
What does CISA's SSVC say?
Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.
How is it classified?
Which compliance frameworks are affected?
This CVE is relevant to:
Frequently Asked Questions
What is CVE-2026-100846?
MONAI versions before 1.5.2 contain an unsafe deserialization flaw in algo_from_pickle (monai/auto3dseg/utils.py), which passes the contents of a .pkl file straight to pickle.loads, so a crafted file executes attacker code with the application's privileges. The package has 126 downstream dependents and 14 other tracked CVEs, and it is widely used in medical imaging pipelines where hosts often hold sensitive patient data. Exploitation likelihood is currently low: EPSS is 0.00292 (top 80th percentile), it is not in CISA KEV, and no public exploit or Nuclei template exists. Still, pickle RCE is trivial to weaponize once a victim loads an attacker-supplied file, and the CVSS of 8.8 reflects full confidentiality, integrity and availability loss. Upgrade to MONAI 1.5.2 or later, and until then block untrusted .pkl files from reaching Auto3DSeg workflows.
Is CVE-2026-100846 actively exploited?
Proof-of-concept exploit code is publicly available for CVE-2026-100846, increasing the risk of exploitation.
How to fix CVE-2026-100846?
1) Upgrade MONAI to 1.5.2 or later (pip install -U monai) and rebuild container images. 2) Until patched, do not call algo_from_pickle on files from untrusted or unverified sources; restrict write access to Auto3DSeg working directories and algorithm template stores. 3) Verify integrity of shared algorithm artifacts with checksums or signatures and keep them in access-controlled storage. 4) Scan pickle artifacts with picklescan or fickling before loading, and prefer non-executable formats where possible. 5) Run training and experimentation jobs with least privilege in sandboxed containers with restricted egress and no ambient cloud credentials. 6) Detect: hunt for unexpected child processes (shell, curl, python -c) spawned by MONAI/Auto3DSeg Python processes, unusual outbound connections from training hosts, and unexpected .pkl files in project directories. 7) Inventory MONAI usage via SBOM and dependency scanning.
What systems are affected by CVE-2026-100846?
This vulnerability affects the following AI/ML architecture patterns: training pipelines, medical imaging ML workflows, model/artifact sharing and registries, Jupyter-based research environments, MLOps automation (AutoML).
What is the CVSS score for CVE-2026-100846?
CVE-2026-100846 has a CVSS v3.1 base score of 8.8 (HIGH). The EPSS exploitation probability is 0.36%.
What is the AI security impact?
Affected AI Architectures
MITRE ATLAS Techniques
AML.T0010.001AI SoftwareAML.T0011.000Unsafe AI ArtifactsAML.T0018.002Embed MalwareAML.T0050Command and Scripting InterpreterCompliance Controls Affected
What are the technical details?
Original Advisory
MONAI before 1.5.2 contains a deserialization of untrusted data vulnerability in the algo_from_pickle function in monai/auto3dseg/utils.py. The function reads a .pkl file and passes its contents to pickle.loads without validating the data source or content. If an application invokes algo_from_pickle on an attacker-supplied pickle file, an object defining __reduce__ is executed during deserialization, resulting in arbitrary code execution in the context of the application.
Exploitation Scenario
An attacker publishes or shares a seemingly legitimate Auto3DSeg algorithm package or a pre-trained pipeline bundle (for example, on a forum, shared drive or a collaborator's repository) containing a malicious .pkl file whose object defines __reduce__ to launch a reverse shell. A researcher or automated pipeline loads it with algo_from_pickle to resume or reuse the algorithm. The payload executes immediately in the context of the MONAI process, exfiltrating imaging datasets and cloud credentials from the GPU workstation, then pivoting to internal storage or PACS systems.
Weaknesses (CWE)
CWE-502 — Deserialization of Untrusted Data: The product deserializes untrusted data without sufficiently ensuring that the resulting data will be valid.
- [Architecture and Design, Implementation] If available, use the signing/sealing features of the programming language to assure that deserialized data has not been tainted. For example, a hash-based message authentication code (HMAC) could be used to ensure that data has not been modified.
- [Implementation] When deserializing data, populate a new object rather than just deserializing. The result is that the data flows through safe input validation and that the functions are safe.
Source: MITRE CWE corpus.
CVSS Vector
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:HReferences
Timeline
Related Vulnerabilities
CVE-2026-487059.3MONAI: NaN/inf DICOM input causes DoS and corrupt inference
Same package: monaiCVE-2025-587558.8MONAI: path traversal allows arbitrary file write
Same package: monaiCVE-2025-587568.8MONAI: unsafe deserialization in CheckpointLoader allows RCE
Same package: monaiCVE-2025-587578.8MONAI: unsafe pickle deserialization RCE in data pipeline
Same package: monaiCVE-2026-1008428.8MONAI: eval sandbox bypass allows code execution
Same package: monai