The vulnerability identified as CVE-2026-25048 affects the mlc-ai xgrammar library, an open-source tool for structured generation. The issue involves uncontrolled recursion triggered by multi-level nested syntax, leading to a segmentation fault and potential denial of service. This flaw is classified under CWE-674 (Uncontrolled Recursion). The vulnerability affects all versions before 0.1.32 and has been addressed by a patch in version 0.1.32. The CVSS 4.0 base score is 8.7, indicating high severity with network attack vector, low attack complexity, no privileges or user interaction required, and high impact on availability. Red Hat has published advisories referencing this CVE, confirming the patch availability.

Exploitation of this vulnerability results in a segmentation fault causing the application to crash, which leads to a denial of service condition. There is no indication of code execution or data compromise beyond service disruption. The impact is high due to the ability to cause application crashes remotely without authentication.

A fix is available in version 0.1.32 of the mlc-ai xgrammar library. Users should upgrade to version 0.1.32 or later to remediate this vulnerability. Red Hat advisories confirm the availability of patches. No additional mitigation steps are specified or required beyond applying the official fix.