CVE-2025-25185: GPT Academic tar.gz symlink flaw enables server file reads

Summary

CVE-2025-25185 affects GPT Academic, an interactive interface for large language models, in version 3.91 and earlier. The software does not properly handle soft links, so an attacker can upload a tar.gz archive containing a soft link that points to a file on the server, and reading the decompressed file exposes that target. The flaw allows attackers to read all files on the server. The weakness is classified as CWE-59, Improper Link Resolution Before File Access.

Mitigation

The source links a patch commit (github.com/binary-husky/gpt_academic/commit/5dffe8627f681d7006cebcba27def038bb691949) and a GitHub security advisory (GHSA-gqp5-wm97-qxcv), but does not state the fixed version or describe the fix in the text.

Related items

• MediumCVE-2026-121936: easyVoice server-side request forgery through openaiBaseUrl in /api/v1/ttsSimilar attack · NVD/CVE Database
• MediumCVE-2026-118628: MLflow get-artifacts server-side request forgery information disclosureSimilar attack · NVD/CVE Database
• HighCVE-2026-118625: MLflow Tracking Server privilege escalation via get_endpointsSimilar attack · NVD/CVE Database
• InfoWhen trust becomes the payload: Inside the threats, etc quarterly reportSimilar attack · Elastic Security Labs
• MediumCritical Langflow flaws enable authentication bypass and RCESimilar attack · CSO Online