ai_security▌

← feed

NOTABLE

HanLP deserialized remote language-model files and extracted model tar archives without path checks

Disclosed together with public exploits and fixed in 2.1.7: CVE-2026-133836 (low) - RNNLanguageModel.load_language_model deserializes a caller-supplied model file in contextual_string_embedding.py, reachable remotely; CVE-2026-133835 (medium) - the uncompress() helper extracts TAR model archives without member-path containment, allowing path traversal.

released 2026-11-21 · ingested 2026-11-21T13:20Z

Sources

Primary source · corroboration

Cite this

APA

AI Security Index. (2026). HanLP deserialized remote language-model files and extracted model tar archives without path checks. AI Security Index. https://github.com/advisories/GHSA-c6gp-rcww-7pq6

BibTeX

@misc{aisec-2026-11-21-hanlp-model-file-deserialization-tar-traversal,
  title  = {HanLP deserialized remote language-model files and extracted model tar archives without path checks},
  year   = {2026},
  howpublished = {\url{https://github.com/advisories/GHSA-c6gp-rcww-7pq6}},
  note   = {AI Security Index item 2026-11-21-hanlp-model-file-deserialization-tar-traversal}
}