Summary
JupyterLab: Cross-site scripting (XSS) in JupyterLab via crafted language package (jupyterlab.json)
Description
A language pack ships a Plural-Forms header saying how the language counts, for example nplurals=2; plural=(n != 1);. JupyterLab turns that string into a function with new Function, so the header gets executed. The check that meant to keep it safe was a regular expression. The regex was anchored at the start but not at the end, so it accepted any string that began with a valid plural rule and ignored everything after it.
A header such as the following passed the check, and the part after the plural rule ran in the JupyterLab page as soon as the first plural string was translated:
nplurals=2; plural=(n > 1); <anything here ran as JavaScript>
Users are affected if all of the following are true:
- they run JupyterLab 3.0.0 through 4.6.3, or an application that bundles it such as Notebook 7;
- a language pack they did not write is installed in the environment; and
- that language is selected, so its catalogue is loaded
An installation using the default English locale loads no catalogue and is not affected.
CVE assignment pending, GitHub CNA is experiencing severe backlog
Workarounds
Use the English locale, which loads no catalogue:
jupyter lab --LabApp.default_locale=en
or the following traitlet:
c.LabApp.default_locale = 'en'
Everyone then sees the interface in English, whatever language they had selected.
To check what is installed instead of switching, report any pack whose header carries more than a plural rule:
python -c "
import re
from jupyterlab_server.translation_utils import get_language_packs, get_language_pack
ok = re.compile(r'\s*nplurals\s*=\s*\d+\s*;\s*plural\s*=[\s\-?|&=!<>+*/%:;n0-9_()]+')
packs, _ = get_language_packs()
for locale in packs:
data, _ = get_language_pack(locale)
for domain, catalog in (data or {}).items():
header = catalog.get('', {}).get('plural_forms')
if header and not ok.fullmatch(header):
print('SUSPECT', locale, domain, repr(header))
"
The command prints nothing when every catalogue is sound. A line of output names the pack to remove.
Impact
The code in the header ran in the JupyterLab page, in the same origin and session as the authenticated user. It could call the Jupyter Server REST API as that user: read and write any file under the server root, start a kernel and run code in it, and open a terminal where terminals are enabled. Nothing had to be clicked; translating one plural string was enough, and that happens during normal use of the interface.
What this changes is who has to be trusted. A language pack is a Python package, and installing one is already a privileged act, so an attacker who can get any package installed has server-side code execution regardless of this issue. The header is different because it is catalogue metadata: it travels with translation content, through the translation pipeline that carries strings from Crowdin into the language packs, and it is reviewed as text rather than as code. Anyone able to change a catalogue, or to publish a pack under a name someone installs, got JavaScript execution in every browser that selected that language.
Note: the impact is much more limited on JupyterLite which typically does not have access to most of the surfaces that this flaw exposes.
Untrusted input is rendered as active markup in a victim's browser, which can run script in their session. Typical impact: session or credential theft, and actions taken as the user.
CVE-2026-102830 has a CVSS score of 6.8 (Medium). The vector is network-reachable, no privileges required, and user interaction required. A CVSS score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether this affects your application depends on whether the vulnerable code is present and reachable in your environment. A fixed version is available (4.6.4, 4.5.11, 0.8.4); upgrading removes the vulnerable code path.
Affected versions
Security releases
Kodem intelligence
Severity tells you how bad this could be in the worst case. It does not tell you whether you are exposed. Exploitability and impact are functions of runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A vulnerable package can sit in your dependency tree and never run.
Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter. Kodem's runtime-powered SCA identifies whether this CVE is reachable in your applications.
Already deployed Kodem?
See it in your environmentNew to Kodem? Get a demo →Remediation advice
JupyterLab v4.6.4 and v4.5.11 contain the patch. The check now has to match the whole header, so a plural rule followed by anything else is rejected and no function is built from it.
JupyterLab 3.x reached end of life and receives no patch. Its users should move to a supported 4.x release.
Users of applications that depend on JupyterLab, such as Notebook v7+, should update jupyterlab package too.
Frequently Asked Questions
- What is CVE-2026-102830? CVE-2026-102830 is a medium-severity cross-site scripting (XSS) vulnerability in jupyterlab (pip), affecting versions >= 4.6.0, <= 4.6.3. It is fixed in 4.6.4, 4.5.11, 0.8.4. Untrusted input is rendered as active markup in a victim's browser, which can run script in their session.
- How severe is CVE-2026-102830? CVE-2026-102830 has a CVSS score of 6.8 (Medium). This score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether it represents real risk in your environment depends on whether the vulnerable code is present and reachable.
- Which packages are affected by CVE-2026-102830?
jupyterlab(pip) (versions >= 4.6.0, <= 4.6.3)jupyterlite-core(pip) (versions <= 0.8.3)
- Is there a fix for CVE-2026-102830? Yes. CVE-2026-102830 is fixed in 4.6.4, 4.5.11, 0.8.4. Upgrade to this version or later.
- Is CVE-2026-102830 exploitable, and should I be worried? Whether CVE-2026-102830 is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
- What actually determines whether CVE-2026-102830 is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.
- How do I fix CVE-2026-102830?
- Upgrade
jupyterlabto 4.6.4 or later - Upgrade
jupyterlabto 4.5.11 or later - Upgrade
jupyterlite-coreto 0.8.4 or later
- Upgrade