Critical Unpatched Flaw Leaves Hugging Face LeRobot Open to Unauthenticated RCE
A serious security vulnerability affecting LeRobot, Hugging Face's open-source robotics platform with almost 24,000 GitHub stars, has been revealed by cybersecurity researchers. This vulnerability might be used to accomplish remote code execution.
The vulnerability in question, CVE-2026-25874 (CVSS score: 9.3), has been characterized as an instance of untrusted data deserialization resulting from the unsafe pickle format.
According to a GitHub advisory for the bug, LeRobot has an unsafe deserialization vulnerability in the async inference pipeline where pickle.loads() is used to deserialize data received via unauthenticated gRPC channels without TLS in the policy server and robot client components.
Sending a forged pickle payload via the SendPolicyInstructions, SendObservations, ...