Impact
The vulnerability resides in the deserialization logic of the AgenticScopeSerializer.fromJson method within LangChain4j’s agentic module. An attacker can supply a crafted JSON payload that the method deserializes unsafely, enabling the instantiation of arbitrary classes. This flaw aligns with CWE-20 (Improper Input Validation) and CWE-502 (Deserialization of Untrusted Data), and it can lead to remote code execution on the host running the application.
Affected Systems
Affected versions are langchain4j 1.5.3-beta10, 1.11.10-beta18, and 1.18.1-beta27. The vendor is the LangChain4j project. All maintained release lines were patched on 29 July 2026, with fixes appearing in 1.5.3-beta11, 1.11.10-beta19, and 1.18.1-beta28.
Risk and Exploitability
The CVSS score of 2.1 indicates low overall severity, yet the vulnerability allows remote exploitation. The EPSS is not available and the issue is not listed in the CISA KEV catalog. Attack complexity is high and exploitation is difficult, but an exploit has been published. Remote attackers would need the application to have AgenticScope persistence enabled and must already have write access to the persistence store. Given the public exploit and the remote nature of the flaw, organizations should consider the risk significant enough to prioritize remediation.
OpenCVE Enrichment