CVE-2026-102830: JupyterLab: language pack Plural-Forms code injection

GHSA-3jqq-pw4j-pqcjMEDIUM
Published September 29, 2026
CISO Take

JupyterLab versions 3.0.0 through 4.5.10 and 4.6.0 through 4.6.3, plus JupyterLite Core 0.8.3 and earlier, evaluate the Plural-Forms header of a third-party language pack with new Function. A weak prefix-only regex lets an attacker append arbitrary JavaScript that runs in the authenticated JupyterLab origin. Where a Jupyter Server is exposed, that script can call authenticated kernel, terminal and file APIs, which gives an attacker file access and code execution on the notebook host. Exploitation is hard to trigger: it needs a victim to install and select a malicious or compromised language pack. The CVSS is 6.8, EPSS is low at 0.26% (83rd percentile), there is no public exploit or Nuclei template, it is not in CISA KEV, and CISA SSVC says TRACK. Still, the package has 1,892 dependents and 41 other CVEs, and data science workstations often hold credentials, datasets and cloud keys. Upgrade to JupyterLab 4.5.11 or 4.6.4 (JupyterLite Core 0.8.4) and allow only vetted language packs.

Sources:NVDEPSSGitHub AdvisoryATLASOpenSSF

What is the risk?

Medium overall. Exploitation needs an attacker to place a crafted language pack in front of a user who then selects that locale (high complexity, user interaction required). The default English locale is not affected. Once triggered, the impact is high for confidentiality and integrity, because the injected code runs with the user's authenticated session and reaches Jupyter Server APIs. There is no known active exploitation, no public PoC and no KEV listing. Risk is highest on multi-user or shared JupyterHub-style deployments, and on hosts with cloud credentials or sensitive training data. JupyterLite exposure is much lower because it usually has no server surface.

How does the attack unfold?

Initial Access
The attacker distributes a malicious or tampered language pack with a crafted Plural-Forms header.
AML.T0010.001
User Execution
The victim installs the pack and selects that locale, so JupyterLab loads the catalogue.
AML.T0011.001
Exploitation
A plural string is translated, new Function evaluates the appended JavaScript, and it runs in the authenticated origin.
AML.T0050
Impact
The script uses Jupyter Server APIs to read or modify files, steal credentials and run code on the host.
AML.T0037

What systems are affected?

PackageEcosystemVulnerable RangePatched
Jupyterpip—No patch
13.5KOpenSSF 5.71.9K dependentsPushed 7d ago54% patched~27d to patchFull package profile →
Jupyterpip>= 4.6.0, <= 4.6.34.6.4
13.5KOpenSSF 5.71.9K dependentsPushed 7d ago54% patched~27d to patchFull package profile →
Jupyterpip<= 0.8.30.8.4
13.5KOpenSSF 5.71.9K dependentsPushed 7d ago54% patched~27d to patchFull package profile →

How severe is it?

CVSS 3.1
6.8 / 10
EPSS
0.3%
chance of exploitation in 30 days
Higher than 18% of all CVEs
Exploitation Status
No known exploitation
Sophistication
Moderate

What is the attack surface?

AVACPRUISCIA
AVNetwork
ACHigh
PRNone
UIRequired
SUnchanged
CHigh
IHigh
ANone

What should I do?

1 step
  1. 1) Upgrade JupyterLab to 4.5.11 or 4.6.4, and JupyterLite Core to 0.8.4. 2) Until patched, stay on the default English locale and remove or do not install third-party language packs (jupyterlab-language-pack-*). 3) Install language packs only from trusted sources, and pin and hash-verify them in internal package mirrors. 4) Run Jupyter Server with token or password authentication, bind it to localhost or a VPN, and give notebook hosts least-privilege credentials. 5) For detection, inventory installed language-pack packages and review their .po or catalogue files for Plural-Forms headers with trailing content after the plural expression. Also watch for unexpected terminal or kernel starts and unusual API calls in Jupyter Server logs.

What does CISA's SSVC say?

DecisionTrack
Exploitationnone
AutomatableNo
Technical Impacttotal

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:

EU AI Act
Art. 15-Accuracy, robustness and cybersecurity
ISO 42001
A.6.2.6-AI system operation and monitoring
NIST AI RMF
MANAGE 2.3-Respond to previously unknown risks
OWASP LLM Top 10
LLM03-Supply Chain

Frequently Asked Questions

What is CVE-2026-102830?

JupyterLab versions 3.0.0 through 4.5.10 and 4.6.0 through 4.6.3, plus JupyterLite Core 0.8.3 and earlier, evaluate the Plural-Forms header of a third-party language pack with new Function. A weak prefix-only regex lets an attacker append arbitrary JavaScript that runs in the authenticated JupyterLab origin. Where a Jupyter Server is exposed, that script can call authenticated kernel, terminal and file APIs, which gives an attacker file access and code execution on the notebook host. Exploitation is hard to trigger: it needs a victim to install and select a malicious or compromised language pack. The CVSS is 6.8, EPSS is low at 0.26% (83rd percentile), there is no public exploit or Nuclei template, it is not in CISA KEV, and CISA SSVC says TRACK. Still, the package has 1,892 dependents and 41 other CVEs, and data science workstations often hold credentials, datasets and cloud keys. Upgrade to JupyterLab 4.5.11 or 4.6.4 (JupyterLite Core 0.8.4) and allow only vetted language packs.

Is CVE-2026-102830 actively exploited?

No confirmed active exploitation of CVE-2026-102830 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-102830?

1) Upgrade JupyterLab to 4.5.11 or 4.6.4, and JupyterLite Core to 0.8.4. 2) Until patched, stay on the default English locale and remove or do not install third-party language packs (jupyterlab-language-pack-*). 3) Install language packs only from trusted sources, and pin and hash-verify them in internal package mirrors. 4) Run Jupyter Server with token or password authentication, bind it to localhost or a VPN, and give notebook hosts least-privilege credentials. 5) For detection, inventory installed language-pack packages and review their .po or catalogue files for Plural-Forms headers with trailing content after the plural expression. Also watch for unexpected terminal or kernel starts and unusual API calls in Jupyter Server logs.

What systems are affected by CVE-2026-102830?

This vulnerability affects the following AI/ML architecture patterns: Interactive notebook environments, Training pipelines, Shared multi-user notebook servers, Browser-based JupyterLite deployments, ML development workstations.

What is the CVSS score for CVE-2026-102830?

CVE-2026-102830 has a CVSS v3.1 base score of 6.8 (MEDIUM). The EPSS exploitation probability is 0.26%.

What is the AI security impact?

Affected AI Architectures

Interactive notebook environmentsTraining pipelinesShared multi-user notebook serversBrowser-based JupyterLite deploymentsML development workstations

MITRE ATLAS Techniques

AML.T0010.001AI Software
AML.T0011.001Malicious Package
AML.T0037Data from Local System
AML.T0050Command and Scripting Interpreter

Compliance Controls Affected

EU AI Act: Art. 15
ISO 42001: A.6.2.6
NIST AI RMF: MANAGE 2.3
OWASP LLM Top 10: LLM03

What are the technical details?

Original Advisory

JupyterLab is an extensible environment for interactive and reproducible computing, based on the Jupyter Notebook Architecture. From JupyterLab 3.0.0 until 4.5.11 and 4.6.4, and in JupyterLite Core 0.8.3 and earlier, the Plural-Forms header in a selected third-party language pack can append JavaScript after a valid plural rule because prefix-only regular-expression validation accepts a matching prefix without requiring the entire header to match. JupyterLab passes the accepted expression to new Function, so loading the catalogue and translating a plural string executes the appended code in the authenticated JupyterLab origin. Where Jupyter Server kernels, terminals, and APIs are exposed, the code can use authenticated server APIs to read or modify files and run code. Impact is much more limited in JupyterLite because it typically lacks most exposed Jupyter Server surfaces. The default English locale is unaffected because it does not load a translation catalogue. This issue is fixed in JupyterLab 4.5.11 and 4.6.4 and JupyterLite Core 0.8.4.

Exploitation Scenario

An attacker publishes or compromises a language pack on PyPI, or tampers with a pack hosted on an internal mirror, and embeds JavaScript after a valid plural rule in its Plural-Forms header. A data scientist installs the pack and switches the interface language. When JupyterLab loads the catalogue and translates a plural string, the payload runs in the authenticated origin. It calls the Jupyter Server contents and terminal APIs to steal cloud credentials and training data, drops a backdoor notebook or startup script, and opens a kernel for further code execution on the ML workstation.

Weaknesses (CWE)

CWE-625 — Permissive Regular Expression: The product uses a regular expression that does not sufficiently restrict the set of allowed values.

  • [Implementation] When applicable, ensure that the regular expression marks beginning and ending string patterns, such as "/^string$/" for Perl.

Source: MITRE CWE corpus.

CVSS Vector

CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:U/C:H/I:H/A:N

Timeline

Published
September 29, 2026
Last Modified
October 1, 2026
First Seen
September 29, 2026

Related Vulnerabilities