JupyterLab Desktop for macOS included the com.apple.security.cs.allow-dyld-environment-variables entitlement before version 4.6.2-1, permitting the dynamic linker to honor DYLD_INSERT_LIBRARIES even with the hardened runtime enabled. An attacker already able to execute code as the current user can load an arbitrary dynamic library into the JupyterLab Desktop process, then inherit the application's TCC permissions. This allows the attacker to read or modify data in protected Documents, Desktop, or Downloads folders without triggering a macOS permission prompt. The vulnerability does not provide initial or remote code execution; it only escalates local privileges over file system resources previously granted to the application.

Apple macOS users running JupyterLab Desktop versions earlier than 4.6.2-1 are affected. The bug does not impact Windows or Linux installations. The issue is confined to the jupyterlab:jupyterlab-desktop product line.

The CVSS score of 5.5 indicates a moderate severity, and the EPSS score is not available, implying limited publicly known exploitation data. Because the flaw requires pre‑existing local code execution and prior TCC grants, the attack vector is local rather than remote. The vulnerability is not listed in the CISA KEV catalog, suggesting it has not yet been widely exploited, but any compromised local session that can execute code would allow access to protected directories.