Critical LangGraph Checkpointer Bugs Enable Remote Code Execution

The article discusses critical vulnerabilities identified in LangGraph's checkpointer, an open-source framework designed for stateful AI agents. Three main vulnerabilities were highlighted: (1) SQL Injection in the SQLite checkpointer (CVE-2025-67644), allowing remote code execution through manipulated SQL queries, (2) Unsafe deserialization through msgpack (CVE-2026-28277), enabling arbitrary code execution via crafted data, and (3) a similar SQL injection vulnerability in the Redis checkpointer (CVE-2026-27022).

Affected users are those self-hosting the framework with exposed `get_state_history()` methods. LangChain, the platform's managing entity, has since patched these vulnerabilities and advised users to update to the latest versions to mitigate risks.