LangChain serialization injection vulnerability CVE-2025-68664

# LangChain serialization injection vulnerability CVE-2025-68664

https://gcdradar.vercel.app/incidents/langchain-serialization-injection-vulnerability-cve-2025-686 · AIR-2025-0388 · closed 2025-12-23

- Lane: Patch now
- Severity: 2 (Disclosed vulnerability)
- Fix status: Fix status unknown
- First reported: 2025-12-23
- AI role: Attack on AI
- Products: LangChain
- CVE ids: CVE-2025-68664
- CVSS: 9.3

## Summary

A serialization injection vulnerability in LangChain's dumps and dumpd functions fails to properly escape dictionaries with lc keys. This allows free-form data to be treated as legitimate LangChain objects during deserialization, enabling secret extraction.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2025-12-23: CVE-2025-68664: LangChain is a framework for building agents and LLM-powered applications. (https://nvd.nist.gov/vuln/detail/CVE-2025-68664)
- 2025-12-23: CVE-2025-68664: LangChain serialization injection vulnerability enables secret extraction in dumps/loads APIs (https://github.com/advisories/GHSA-c67j-w6g6-q2cm)

## Sources

- 2025-12-23 GitHub Advisory Database: CVE-2025-68664: LangChain serialization injection vulnerability enables secret extraction in dumps/loads APIs https://github.com/advisories/GHSA-c67j-w6g6-q2cm
- 2025-12-23 NVD: CVE-2025-68664: LangChain is a framework for building agents and LLM-powered applications. https://nvd.nist.gov/vuln/detail/CVE-2025-68664

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.