CVE-2026-25874 is a critical vulnerability in Hugging Face's LeRobot product involving unsafe deserialization of untrusted data. The async inference pipeline uses Python's pickle.loads() to deserialize data received over unauthenticated and unencrypted gRPC channels in the policy server and robot client components. This unsafe deserialization allows an unauthenticated remote attacker to execute arbitrary code by sending maliciously crafted pickle payloads through the SendPolicyInstructions, SendObservations, or GetActions gRPC calls. The vulnerability affects version 0 of LeRobot and has a CVSS 4.0 base score of 9.3, reflecting high attack vector and impact metrics. No patch or official remediation level has been published as of the vulnerability disclosure date.

An unauthenticated attacker with network access to the affected LeRobot components can achieve arbitrary code execution on both the server and client by exploiting unsafe deserialization over unauthenticated gRPC channels. This can lead to full compromise of the affected systems running LeRobot version 0. The vulnerability is critical due to the lack of authentication, encryption, and the use of insecure deserialization methods, allowing remote code execution without user interaction or privileges.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until an official fix is released, users should consider restricting network access to the affected gRPC services, implementing network-level authentication or encryption (e.g., TLS), or disabling the vulnerable async inference pipeline components if feasible. Avoid exposing the policy server and robot client components to untrusted networks. Monitor vendor communications for updates on patches or official mitigations.