CVE-2026-102938

A flaw was found in virtualenv, a tool for creating isolated Python virtual environments that is bundled in a number of Red Hat offerings, including Red Hat Enterprise Linux, Red Hat OpenShift AI, and other layered products and containers, as well as in Fedora and EPEL community builds. Red Hat rates this flaw as having Moderate impact. virtualenv writes prompt values verbatim into the line-oriented pyvenv.cfg file, while its reader parses the file with str.splitlines() and accepts the last value for duplicate keys. An attacker who can influence prompt input -- the --prompt command-line option, the VIRTUALENV_PROMPT environment variable, or a configuration-derived value such as a CI job templating an untrusted branch name -- can inject line boundaries and additional configuration keys, including home, into the generated pyvenv.cfg. Because tooling reads home to locate the base interpreter, a consumer that trusts the file can be redirected to an attacker-selected interpreter, potentially resulting in arbitrary code execution or environment tampering when the virtual environment is used. Exploitation requires the prompt value to come from outside the operator's trust boundary: where the operator supplies the prompt directly, they already control the command line, and the effect is corruption of their own environment rather than privilege gain. This is why Red Hat assesses the impact as Moderate, while the upstream project published the advisory with low severity. Fixes for affected Red Hat products and streams are being pursued by the responsible component and product teams under delegated resolution; some community and out-of-support-scope streams will not receive an update.