CVE-2026-112145: Use of Externally-Controlled Input to Select Classes or Code in speaches-ai speaches

The vulnerability CVE-2026-112145 affects the speaches-ai speaches project up to version 0.9.0-rc.3. It resides in the PyannoteDiarizationModelManager._load_fn function within src/speaches/executors/pyannote_diarization.py. The flaw allows an attacker to manipulate the 'model' argument, which is used as externally-controlled input to select classes or code, creating a risk of unintended code execution. This vulnerability can be exploited remotely. It was introduced in a commit identified as 870e1e16a73452496e99d49f759eabd692a8e0e0, merged as part of PR #628. The project was notified early but has not yet issued a fix or official response.

An attacker can remotely exploit this vulnerability by manipulating input to influence class or code selection, potentially leading to unauthorized code execution or other unintended behavior. The CVSS 4.0 base score is 6.9 (medium), reflecting network attack vector, low complexity, no privileges or user interaction required, but limited confidentiality, integrity, and availability impacts.

No official fix or patch is currently available, and the vendor has not responded to the issue report. Users should monitor the vendor's communications for updates. Until a patch is released, avoid exposing the vulnerable component to untrusted inputs or networks where possible.