Deserialization of Untrusted Data in LangChain - #VU130244
Published: May 6, 2026
LangChain
LangChain
Description
The vulnerability allows a remote attacker to disclose sensitive information and manipulate application behavior.
The vulnerability exists due to unsafe deserialization in the load() deserialization logic when processing untrusted structured input that is later deserialized from LangChain run data. A remote attacker can submit crafted LangChain serialized constructor dictionaries to disclose sensitive information and manipulate application behavior.
Applications are exposed only if untrusted structured input is preserved in run inputs or outputs and later reaches affected runtime surfaces such as RunnableWithMessageHistory, astream_log(), or astream_events(version="v1").