UpTrain, an open‑source AI evaluation platform, has a flaw in its /add_prompts HTTP endpoint that lets authenticated users inject data into the checks and metadata parameters. The input is passed to python code without proper sanitization, which allows an attacker to execute arbitrary code on the host system, typically the docker container in which UpTrain runs. This results in full compromise of the container and potential lateral movement inside the host environment.

The vulnerability exists in UpTrain version 0.7.1 and all earlier releases. Users running these versions on any platform that follows the standard docker-based deployment are affected.

The CVSS score for this issue is 7.4, indicating moderate to high severity. EPSS data is not available and the vulnerability is not currently listed in the CISA KEV catalog. The likely attack vector requires a valid authenticated session to access the /add_prompts endpoint, making it a privilege‑related RCE. An attacker who can authenticate could fully compromise the container and, in many deployment scenarios, the underlying host system.