CVE-2025-27770 describes an injection vulnerability (CWE-74) in the open-source UpTrain platform used for evaluating generative AI applications. Specifically, the /create_project endpoint improperly neutralizes special elements in the checks and metadata parameters, allowing authenticated users to execute arbitrary code remotely in the host environment. This vulnerability affects UpTrain version 0.7.1 and prior. There is no known patch or official remediation at the time of publication.

An attacker with valid authentication to UpTrain can exploit this vulnerability to execute arbitrary code on the host running the UpTrain service, which is commonly a Docker container. This could lead to full compromise of the container environment and potentially the underlying host system depending on deployment configurations.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until a fix is available, restrict access to authenticated users carefully and consider isolating the UpTrain environment to limit potential impact.