Executive brief
JupyterLab is an interactive computing environment used by data scientists and researchers to write and run code. A security flaw in the extension manager allows authenticated users to bypass restrictions on which software extensions can be installed. In shared or multi-tenant environments, such as a university JupyterHub, this could allow a user to escalate their privileges, steal data, or gain persistent access to the underlying server infrastructure.
Technical details
JupyterLab versions 4.0.0 through 4.5.6 contain a vulnerability where the 'allowed_extensions_uris' configuration is not correctly enforced. The PyPI Extension Manager fails to restrict package installation to the default PyPI index or the specified allow-list, leading to potential argument injection (CWE-88) or client-side enforcement issues (CWE-602). An authenticated attacker in a multi-tenant or shared deployment (like JupyterHub) can exploit this to install malicious extensions, resulting in privilege escalation, lateral movement, or persistent server compromise. The issue is fixed in version 4.5.7; a workaround is to set the extension manager to 'readonly' mode.
Affected products
- Project Jupyter JupyterLab 4.0.0 to 4.5.6
Timeline
- 2026-04-29: patched: Version 4.5.7 released with security fixes.
- 2026-04-30: advisory: GitHub security advisory GHSA-37w4-hwhx-4rc4 published.
- 2026-05-13: disclosed: CVE-2026-42266 published to NVD.