Serialization Vulnerability in LangChain Framework by LangChain AI
CVE-2026-44843

8.2HIGH

Key Information:

Vendor

Langchain-ai

Status
Langchain
Vendor
CVE Published:
26 May 2026

What is CVE-2026-44843?

The LangChain framework, which facilitates the development of agents and LLM-powered applications, has a serialization vulnerability related to deserializing run inputs and outputs. Versions prior to 0.3.85 and 1.3.3 contain outdated runtime code paths that utilize overly broad object allowlists. These paths allow the use of the load() function with permitted object types set to 'all'. Although this does not permit arbitrary Python object deserialization, it enables maliciously crafted LangChain serialized constructor dictionaries to instantiate classes with untrusted arguments. This poses a significant risk, as it allows attackers to manipulate trusted runtime processes. The issue has been addressed in the latest updates.

Affected Version(s)

langchain < 0.3.85 < 0.3.85

langchain >= 1.0.0a1, < 1.3.3 < 1.0.0a1, 1.3.3

References

https://github.com/langchain-ai/langchain/security/adviso...
x_refsource_CONFIRM

CVSS V3.1

Score:
8.2
Severity:
HIGH
Confidentiality:
High
Integrity:
Low
Availability:
High
Attack Vector:
Network
Attack Complexity:
Low
Privileges Required:
None
User Interaction:
None
Scope:
Unchanged

Timeline

  • Vulnerability published

  • Vulnerability Reserved

.