Multiple vulnerabilities affect the Natural Language Toolkit (NLTK) Python library across a wide severity spectrum, ranging from critical sandbox bypasses (CVSS 9.8) to high-severity denial-of-service and untrusted search path flaws (CVSS 7.3–7.5) and medium-severity parsing, recursion, and path issues (CVSS 5.3–6.5). Several additional vulnerabilities lack published CVSS scores or have scores not stated in the advisory dataset. Attackers processing untrusted text, corpus files, or grammar/tree structures in NLTK-backed services risk code execution, unauthorized file access, or service disruption. Exposure is highest where applications pass user-supplied inputs directly into tokenizers, parsers, corpus readers, or external execution helpers.
CVE-2026-129234: This high-severity (CVSS 7.3) untrusted search path flaw affects NLTK search and execution helpers. Attackers who can influence search paths or local resources can redirect execution or data loading. Successful exploitation risks arbitrary code or unintended resource execution, posing moderate-to-high risk in multi-user or shared environments.[/subscribe_to_unlock_form]
Multiple vulnerabilities affect the Natural Language Toolkit (NLTK) Python library across a wide severity spectrum, ranging from critical sandbox bypasses (CVSS 9.8) to high-severity denial-of-service and untrusted search path flaws (CVSS 7.3–7.5) and medium-severity parsing, recursion, and path issues (CVSS 5.3–6.5). Several additional vulnerabilities lack published CVSS scores or have scores not stated in the advisory dataset. Attackers processing untrusted text, corpus files, or grammar/tree structures in NLTK-backed services risk code execution, unauthorized file access, or service disruption. Exposure is highest where applications pass user-supplied inputs directly into tokenizers, parsers, corpus readers, or external execution helpers.
CVE-2026-129234: This high-severity (CVSS 7.3) untrusted search path flaw affects NLTK search and execution helpers. Attackers who can influence search paths or local resources can redirect execution or data loading. Successful exploitation risks arbitrary code or unintended resource execution, posing moderate-to-high risk in multi-user or shared environments.[emaillocker id="1283"]
CVE-2026-129233: Rated medium severity (CVSS 5.3), this vulnerability impacts NLTK input handling and processing routines. Malformed or adversarial inputs can trigger unexpected behavior or service degradation. Exploitation impact is primarily limited to availability or minor data integrity issues, with lower risk when inputs are strictly validated.
CVE-2026-63312: With a CVSS score not stated, this vulnerability involves a path security bypass in NLTK StreamBackedCorpusView. Attackers supplying crafted paths or corpus references can read arbitrary local files. Exploitation risk is elevated wherever untrusted corpus paths or file selectors are exposed to external users.
CVE-2026-129232: Assigned a medium CVSS score of 5.3, this flaw affects NLTK text or structure processing components. Specially crafted inputs can cause abnormal processing states or resource exhaustion. Exploitation risk remains moderate and depends on whether untrusted data reaches vulnerable functions.
CVE-2026-81723: Lacking a stated CVSS score, this issue affects NLTK XML corpus parsing components. Malformed XML corpus files can trigger denial-of-service conditions during parsing. Exploitation risk increases in pipelines that ingest unverified external XML corpora.
CVE-2026-81725: With no CVSS score stated, this regular expression denial-of-service flaw resides in NLTK Pl196xCorpusReader. Crafted corpus input can induce excessive regex evaluation time and stall processing. Exploitation risk is moderate when processing untrusted corpus data without timeouts or input size limits.
CVE-2026-129231: Rated medium severity (CVSS 5.3), this vulnerability affects NLTK internal parsing or evaluation helpers. Adversarial inputs can disrupt normal execution flow or degrade performance. Exploitation risk is generally contained to service availability impacts under untrusted input exposure.
CVE-2026-129230: Carrying a medium CVSS score of 5.5, this issue impacts NLTK data handling or token processing routines. Untrusted inputs can lead to unexpected execution paths or availability disruption. Exploitation risk is moderate in automated text-processing services handling external payloads.
CVE-2026-129229: This high-severity (CVSS 7.5) compile-time regular expression denial-of-service vulnerability targets NLTK search and tokenizer sinks. Quadratic regex compilation bypasses match-only timeouts and can lock application threads. Exploitation risk is high for web or API endpoints exposing search or tokenization to untrusted strings.
CVE-2026-80205: With a CVSS score not stated, this regular expression denial-of-service flaw affects Text.findall() via unvalidated regular expressions. Attackers supplying malicious patterns or inputs can cause severe CPU and time consumption. Exploitation risk is notable when user-controlled regexes or large texts are passed to Text.findall().
CVE-2026-129228: Assigned a medium CVSS score of 5.3, this vulnerability affects NLTK processing modules handling structured or unstructured text. Crafted inputs can trigger processing inefficiencies or stability issues. Exploitation risk is moderate where external input reaches affected routines directly.
CVE-2026-129227: Rated medium severity (CVSS 5.3), this uncontrolled recursion flaw in Tree.fromlist triggers denial of service via deeply nested list inputs. Deep nesting can exhaust call stack resources and crash Python worker processes. Exploitation risk is elevated when untrusted nested structures are converted into parse trees without depth limits.
CVE-2026-129226: Carrying a medium CVSS score of 5.3, this issue affects NLTK parsing or tree/sequence handling routines. Malformed or deeply structured inputs can degrade service responsiveness or stability. Exploitation risk depends on whether untrusted input is passed into vulnerable parsing paths.
CVE-2026-129225: With a medium CVSS score of 5.3, this vulnerability impacts NLTK text or corpus processing components. Crafted inputs can cause abnormal resource consumption or application errors. Exploitation risk is moderate when processing unverified external text or datasets.
CVE-2026-129224: Assigned a medium CVSS score of 5.3, this flaw affects NLTK internal processing logic on crafted inputs. Adversarially formatted data can impair availability or cause unexpected exceptions. Exploitation risk remains moderate under standard input validation controls.
CVE-2026-129223: Rated high severity (CVSS 7.5), this recursion flaw affects NLTK recursive parsing or evaluation routines. Unbounded or deeply nested inputs can exhaust stack capacity and cause denial of service. Exploitation risk is high for services accepting arbitrary nested inputs without depth constraints.
CVE-2026-129222: This high-severity (CVSS 7.5) vulnerability impacts NLTK processing routines capable of triggering severe resource or service disruption. Maliciously crafted inputs can reliably degrade or halt affected components. Exploitation risk is high when untrusted payloads reach vulnerable endpoints directly.
CVE-2026-129221: Carrying a medium CVSS score of 5.4, this issue allows ARFF_Formatter in nltk.classify.weka to inject Weka ARFF schema directives via attacker-controlled class labels. Schema injection can corrupt dataset definitions or alter downstream classification behavior. Exploitation risk is moderate when exporting untrusted labels into ARFF format.
CVE-2026-128140: Rated medium severity (CVSS 6.5), this vulnerability affects NLTK data or file handling components. Crafted inputs or paths can lead to unauthorized access or processing flaws depending on deployment context. Exploitation risk is moderate where external input influences file or data selection.
CVE-2026-128139: Assigned a high CVSS score of 7.5, this vulnerability affects NLTK execution or processing components with significant impact potential. Malicious inputs can trigger severe service disruption or unintended behavior. Exploitation risk is high in exposed services processing untrusted inputs.
CVE-2026-128138: This critical-severity (CVSS 9.8) vulnerability represents an incomplete fix enabling a JVM/JAR sandbox bypass through unvalidated launcher tokens in java() command arguments. Attackers controlling launcher arguments can escape sandbox restrictions and execute arbitrary code. Exploitation risk is critical wherever NLTK invokes external Java tools with untrusted parameters.
CVE-2026-12841: With a CVSS score not stated, this vulnerability affects NLTK components handling external or structured inputs. Improper handling of crafted data can cause security or stability impacts. Exploitation risk depends on exposure of affected functions to untrusted callers.
CVE-2026-128137: Rated medium severity (CVSS 5.5), this flaw impacts NLTK processing or resource management routines. Adversarial inputs can cause denial-of-service conditions or unexpected errors. Exploitation risk is moderate when processing unvalidated external text or structures.
CVE-2026-128136: Carrying a medium CVSS score of 6.1, this vulnerability affects NLTK input handling or response/output generation paths. Crafted inputs can lead to information exposure, injection, or processing anomalies. Exploitation risk is moderate in web or API workflows handling untrusted data.
CVE-2026-128135: This high-severity vulnerability affects NLTK local file or resource handling with notable integrity and availability impact. Local or untrusted inputs can overwrite or disrupt data and service availability. Exploitation risk is significant in shared or multi-tenant environments running NLTK.
Addressing this broad set of NLTK flaws requires prompt dependency updates, strict input validation, and sandboxing around text, corpus, and external command execution. Organizations should prioritize patching high- and critical-severity paths while enforcing timeouts, recursion limits, and path controls across all NLP pipelines.
We recommend you to update nltk to version 3.10.4, 3.10.0, or 3.10.3 depending on your installed branch.
The following reports contain further technical details: