Agent Frameworks

LangChain 0.3.29 patches deserialization and load() security flaws

Two critical hardening fixes to prevent untrusted manifest attacks

Deep Dive

LangChain released version 0.3.29 with fixes that restrict deserialization in langchain.storage._lc_store and harden load() against untrusted manifests.

Key Points
  • Restricts unsafe deserialization in langchain.storage._lc_store to block object injection
  • Hardens load() against untrusted manifests in both langchain and langchain-core
  • No feature additions; purely a security patch requiring upgrade for production use

Why It Matters

Essential security patch for LangChain users; unpatched instances risk RCE via untrusted serialized data.

📬 Get the top 10 AI stories daily