The FastAPI AI inference server used by EnergizeLab’s Eilik and EilikO robots allows uploading of AI skill files without requiring authentication and without validating the file path. An attacker can exploit this by appending path traversal sequences to the upload name, causing arbitrary files to be written on the server’s filesystem. If the attacker writes malicious code or alters configuration files used by the AI inference engine, they can potentially execute arbitrary code, alter model behavior, or tamper with robot operations, compromising confidentiality, integrity, and availability of the system.

This vulnerability affects the EnergizeLab FastAPI AI inference and profiling services, specifically the Eilik and EilikO AI skill upload endpoints. No specific version numbers are disclosed in the advisory, so any deployment of these components that has not applied the latest vendor update is potentially vulnerable.

The vulnerability carries a CVSS score of 9.1, indicating a high severity level. The EPSS score is not available, and the issue is not listed in CISA’s KEV catalog, but the absence of an authentication requirement and the ability to write arbitrary files mean that a remote attacker can exploit this from any network position that can reach the upload endpoint. The primary attack vector is an unauthenticated remote web request, thus the risk of exploitation remains significant.