LangGraph Flaw Chain Exposes Self-Hosted AI Agents to Remote Code Execution

Security researchers disclosed and LangGraph maintainers have patched three issues in LangGraph that can be chained into remote code execution for self-hosted deployments using the SQLite or Redis checkpointer. The affected vulnerabilities are CVE-2025-67644, CVE-2026-28277, and CVE-2026-27022, which respectively involve SQL injection in SQLite checkpoints, unsafe msgpack deserialization, and a RediSearch query injection in @langchain/langgraph-checkpoint-redis. This matters because the chain can allow attackers to turn tampered checkpoint data into server-side code execution, potentially exposing runtime secrets or other connected systems.