The vulnerability arises from deserialization of untrusted data within the run_in_subprocess_wrapper_func function of GPT Academic. Because user-supplied input is not validated before deserialization, a remote attacker can craft data that, when processed, is executed as code with root privileges. This flaw allows arbitrary code execution and has no authentication requirement.

GPT Academic's GPT Academic software, version 3.91, is affected. No other versions are listed as vulnerable in the available data.

The CVSS score of 9.8 indicates a critical severity, while an EPSS score of 2% suggests a modest but non-negligible likelihood of exploitation. The vulnerability is not yet listed in the CISA KEV catalog. Based on the description, the likely attack vector is remote, network-based delivery of crafted data to the vulnerable function, enabling control over the execution context without authentication.