NLTK arbitrary code execution via Java launcher arguments
Affects nltk 3.10.3 in applications that let attackers influence the java() cmd argument or StanfordSegmenter java_class. CVSS rates it Critical (9.8). Fixed in 3.10.4.
- Exploited
- Proof of concept public, per CISA
- Fix
- Fixed in 3.10.4
- Who is exposed
- Not reported
Summary
NLTK, the Natural Language Toolkit Python library, has an incomplete fix for an earlier Java sandbox bypass. Supplying -jar or an @argfile through its Java launcher helpers lets an attacker execute Java bytecode of their choice. An attacker can do this remotely, without logging in and without any user action.
Written by a model
Original description
NLTK (Natural Language Toolkit) is a suite of open source Python modules, data sets, and tutorials supporting research and development in Natural Language Processing. From 3.10.3 until 3.10.4, the NLTK 3.10.3 JVM hardening validates Java launcher tokens in the options channel but appends the caller-controlled cmd sequence without requiring its first token to be a main class. Supplying -jar or an @argfile through nltk.internals.java or StanfordSegmenter.java_class can therefore execute attacker-selected Java bytecode without the unsafe-JAR opt-in. This issue is fixed in version 3.10.4.
Who should check
Teams running nltk 3.10.3 whose code passes untrusted or attacker-influenced values into nltk.internals.java cmd or StanfordSegmenter java_class.
“Practical exploitability depends on an embedding application exposing attacker influence over one of these public argument paths”
What to do
Upgrade to a fixed version: 3.10.4.
Check your systems with an agent
Paste this into a coding or ops agent. It checks your systems read-only and reports a table per asset.
Check whether my systems are affected by CVE-2026-128138 (NLTK incomplete fix: additional JVM/JAR sandbox bypass through unvalidated launcher tokens in java() command arguments), using the cve.now data below.
Rules: work read-only. Do not run exploit code, proof-of-concept scripts or scanners that send attack payloads, and do not change any configuration, package, image or deployment. Report findings only.
What cve.now has for CVE-2026-128138 (entry last updated 2026-11-19):
Affected:
- nltk nltk: from 3.10.3, fixed in 3.10.4
Fixed versions: 3.10.4.
Conditions (all must be true for the bug to be hit):
- Affects nltk 3.10.3, the version with the JVM hardening. (source: https://github.com/nltk/nltk/security/advisories/GHSA-x8q7-6c7m-52gw)
- An embedding application must expose attacker influence over the java() cmd argument or StanfordSegmenter java_class. (source: https://github.com/nltk/nltk/security/advisories/GHSA-x8q7-6c7m-52gw)
Setup: unknown whether the bug needs a non-default feature or setting. Check the advisory.
Attacker needs: network access, no login, no user action.
CISA KEV: CVE-2026-128138 is not listed as of cve.now's last read.
This KEV status is for CVE-2026-128138 only. If you also check another CVE, look up its own KEV status; never assume it from this one.
How to check:
- Installed software: check what is actually running, not what a repo declares. Read the version from each host, container image and Kubernetes node (the binary's version output, the OS package database, image layers), and check whether the conditions above hold there. Distribution packages may carry a backported fix under an older version number; confirm against the distribution's own advisory.
Verify against the linked advisories before concluding, and cite the date you read each one:
- https://github.com/nltk/nltk/security/advisories/GHSA-x8q7-6c7m-52gw
- https://github.com/nltk/nltk/pull/3771
- https://github.com/nltk/nltk/commit/b36212795a522d2b7b176cf0e0d9d8da489ba6e6
- https://github.com/nltk/nltk/releases/tag/v3.10.4
Do not conclude an asset is unaffected unless you checked it and the evidence covers both its version and the conditions above. If you could not check something, mark it unknown and say what is missing.
Report a table with one row per asset and these columns: Asset | Version or build | Status (affected, fixed, not applicable, unknown) | Compensating controls | Exposure (internet-facing, internal, none) | Evidence | Next action.
Entry: https://cve.now/CVE-2026-128138
For follow-up questions, use the cve.now MCP server at https://mcp.cve.now/mcp (no authentication), for example execute with name cve.get and arguments {"id": "CVE-2026-128138"}.Affected versions
- nltk nltkFrom 3.10.3 · fixed in 3.10.4
cve.now Index
- Exploitation15/40CISA rates a proof of concept as public
- Exposure20/25No login or user action needed, CISA rates it automatable · Unknown: Advisories don't say whether default setups are affected
- Reach?/20Unknown. No tracked technology or package matched
- Severity10/10CVSS 9.8, total technical impact per CISA
- No patch0/5No signal
What would change the score
- Would score about 75 if added to CISA KEV
- Would score about 70 if CISA's SSVC rates exploitation active
- Would score about 50 if advisories confirm default setups are exposed
Timeline
- CVE published
- First seen by cve.now
- Posted on X by cve.now
- CISA SSVC assessment
Exploitation: poc · Automatable: yes · Technical impact: total
- Record last updated
Evidence 6
| Kind | Source |
|---|---|
| Vendor advisory 1Establishes: Affected and fixed versions. | |
| Advisory | GitHub · nltk/nltk · Advisory GHSA-x8q7-6c7m-52gw |
| CVE record 2Establishes: The identifier and description. | |
| Record | CVE List · CVE-2026-128138 |
| Record | NVD · CVE-2026-128138 |
| Other links 3 | |
| Repository | GitHub · nltk/nltk · Pull request #3771 |
| Repository | GitHub · nltk/nltk · Commit b3621279 |
| Repository | GitHub · nltk/nltk · Release v3.10.4 |
Questions about CVE-2026-128138
Is CVE-2026-128138 being exploited?
It is not in CISA's Known Exploited Vulnerabilities catalog, and CISA's SSVC assessment does not mark exploitation as active. That does not mean it has not been attacked: vendors and researchers can report attacks before CISA lists them. CISA's SSVC assessment notes a proof of concept exists. No public exploit code is tracked.
Source: cve.now feed data
What does this mean?
An earlier fix for a JVM/JAR sandbox bypass in NLTK left some launcher tokens in java() command arguments unvalidated. Attacker-selected Java bytecode can then run with the privileges of the process using NLTK.
Source: github.com
Which versions are affected by CVE-2026-128138?
nltk: from 3.10.3 before 3.10.4.
Source: cve.now feed data
Am I affected, and how do I check?
Only nltk 3.10.3 is confirmed affected, and only where an embedding application exposes attacker influence over the java() cmd argument or StanfordSegmenter java_class. To check, look at the installed nltk version (for example with pip show nltk) and whether those arguments can be influenced from outside the application.
Source: github.com
How do I fix CVE-2026-128138?
Upgrade nltk to 3.10.4, the listed fixed version.
Source: github.com
Related
- CVE-2026-128134NLTK dot2img runs Graphviz 'dot' without the trusted-exec chokepoint, enabling local code execution via a writable PATH element
- CVE-2026-129223Natural Language Toolkit (NLTK): Uncontrolled recursion (crash DoS) in Synset.tree()/acyclic_tree() via WordNet relation traversal
- CVE-2026-128139NLTK: XMLCorpusView comment and CDATA parsing permits quadratic CPU denial of service
- CVE-2026-129222Natural Language Toolkit (NLTK): Uncontrolled recursion (crash DoS) in DependencyGraph.tree()/triples()
- CVE-2026-129229NLTK: Compile-time regex DoS in NLTK search/tokenizer sinks: quadratic regex-engine compile bypasses the match-only timeout
- CVE-2026-129234NLTK: Untrusted search path in prover9/mace4 config allows local code execution via relative binary_location
No corrections so far. Something wrong on this page? Report an error.