CVE-2026-51857Published Sep 30, 2026 · Sep 30Updated Oct 2
In camel-ai camel 0.2.91a1, v0.2.91a2 and v0.2.91a3, CodeExecutionToolkit can run model-produced Python code through…
Critical9.8
In camel-ai camel 0.2.91a1, v0.2.91a2, and v0.2.91a3, CodeExecutionToolkit can run model-produced Python code through SubprocessInterpreter without an approval boundary. This can result in execution of that code; the input does not specify network access, authentication requirements, or user interaction.
Engineers running camel-ai camel 0.2.91a1, v0.2.91a2, or v0.2.91a3 should check their deployments.