Threat
LangChain Serialization Injection Vulnerabilities Enable Secret Extraction
Publication date unknown · Discovered mallory.ai
Page published
Publication date unknown · First observed: 18 Oct 2026
Coverage timeline
Single-source incident — one report is available.
Why it matters
LangChain applications that deserialize untrusted or prompt-injected data may turn attacker-controlled content into secret access or internal object reconstruction, making upgrades to the patched packages a priority.
LangChain serialization injection vulnerabilities CVE-2025-68665 in JavaScript and CVE-2025-68664 in Python allow attacker-controlled structures containing the reserved `lc` key to be reconstructed as trusted internal objects, potentially exposing environment secrets or manipulating application behavior. Patched releases are @langchain/core 0.3.80 and 1.1.8, langchain 0.3.37 and 1.2.3, and Python langchain-core 0.3.81 and 1.2.5.