CVE-2026-102832: JupyterLab command link trust prompt marks wrong document

CVE-2026-102832 · Severity: medium · CVSS 4.2 · Published 2026-10-13

JupyterLab is an interactive notebook environment used for data analysis and research. When a user clicks a button in a Markdown file or notebook to run a command, a trust confirmation dialog can incorrectly mark a different document as trusted if another document becomes active while the dialog is open. This allows buttons in the wrong document to execute without user confirmation, defeating the security protection that was added to prevent uncontrolled command execution.

Technologies: JupyterLab, jupyterlab (PyPI), jupyterlite-core (PyPI), notebook (PyPI). Vendors: Jupyter, JupyterLab, PyPI.

Quick answers

Is CVE-2026-102832 exploited in the wild?
No. Junglewise Threat Intelligence has no report of CVE-2026-102832 being exploited in the wild.
How severe is CVE-2026-102832?
Medium severity, with a CVSS score of 4.2, published 13 October 2026.
Which products and versions are affected?
Jupyter Notebook 7.5 and later (which bundles JupyterLab); JupyterLab 4.5.7 through 4.6.3; jupyterlab (PyPI) >= 4.6.0, <= 4.6.3, ECOSYSTEM:introduced=4.6.0,fixed=4.6.4; jupyterlite-core (PyPI) >= 0.7.5, <= 0.8.3, ECOSYSTEM:introduced=0.7.5,fixed=0.8.4; and 1 more.
Is there a fix for CVE-2026-102832?
Yes. jupyterlab (PyPI): upgrade to 4.6.4 or later; jupyterlite-core (PyPI): upgrade to 0.8.4 or later; notebook (PyPI): upgrade to 7.6.3 or later. Until then: Disable command linker by setting allowCommandLinker to false in JupyterLab configuration, which prevents command execution from buttons entirely. Alternatively, users can select "Run Once" instead of "Trust" when prompted, which executes the command without marking any document as trusted. Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.
How is CVE-2026-102832 attacked?
It is attacked over the network or with local access, in a web application.
Am I affected by CVE-2026-102832?
Check whether you run Jupyter Notebook, JupyterLab, jupyterlab (PyPI) or one of the others listed at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

How to fix

  • jupyterlab (PyPI): upgrade to 4.6.4 or later
    pip install "jupyterlab>=4.6.4"
  • jupyterlite-core (PyPI): upgrade to 0.8.4 or later
    pip install "jupyterlite-core>=0.8.4"
  • notebook (PyPI): upgrade to 7.6.3 or later
    pip install "notebook>=7.6.3"

Disable command linker by setting allowCommandLinker to false in JupyterLab configuration, which prevents command execution from buttons entirely. Alternatively, users can select "Run Once" instead of "Trust" when prompted, which executes the command without marking any document as trusted.

Technical details

A race condition in the trust command handlers (markdownviewer:trust and notebook:trust) reads the currently active document when the user answers the prompt, rather than when they click the button. If a different document becomes current while the dialog is open, that document is marked trusted instead of the one the user intended. For notebooks, this triggers a file write with cryptographic signing, making the trust persistent across sessions.

Affected products

  • JupyterLab JupyterLab 4.5.7 through 4.6.3
  • Jupyter Notebook 7.5 and later (which bundles JupyterLab)

Identifiers

  • CVE-2026-102832
  • GHSA-jwrc-gm9j-263p

Timeline

  • 2026-10-15: disclosed
  • 2026-10-15: patched

This page as Markdown. Free to reuse under CC BY 4.0 with a link back.