Hydra, a framework for configuring complex applications, has a vulnerability in the hydra.utils.instantiate function that allows attacker-controlled configuration values to be resolved and executed. The flaw arises from the _resolve_target() logic, which accepts target names and arguments from untrusted inputs and calls the corresponding Python objects. This leads to arbitrary code execution in the process, enabling an attacker to read or modify files, leak credentials, or terminate the application. The weakness is classified under CWE-470 (Use of a Dangerous Class) and CWE-94 (Improper Control of Generation of Code).

The issue affects all releases of Hydra before version 1.3.4 produced by Facebook Research. Any application, library, or command‑line flow that incorporates Hydra and feeds untrusted configuration, CLI overrides, or model metadata into hydra.utils.instantiate() is potentially vulnerable.

The CVSS score for this vulnerability is 7.8, indicating a high severity. The EPSS score is not available, but the flaw’s high severity suggests that exploitation could be attempted once the attacker can supply arbitrary configuration. The vulnerability is not listed in the CISA KEV catalog. The likely attack vector is through untrusted configuration files or command‑line overrides that are processed by hydra.utils.instantiate, allowing an attacker to specify dangerous callables and arguments.