Executive brief
The Drupal AI module, which integrates artificial intelligence and language models into websites, contains a security flaw that could allow unauthorized access to certain administrative actions. If a site is configured to allow non-privileged users to interact with AI agents, an attacker could potentially trigger system actions they are not authorized to perform. This could lead to unauthorized data modifications or exposure of internal site functions.
Technical details
A missing authorization vulnerability (CWE-862) exists in the Drupal AI (Artificial Intelligence) module. The flaw stems from incorrect access validation in certain Drupal core actions exposed as agent tools, as well as missing access-level definitions for some core actions. An attacker with permission to communicate with an affected AI agent can perform 'forceful browsing' to execute these actions without proper authorization. Exploitation is mitigated by the requirement that the site must be specifically configured to expose these tools to non-privileged users. The issue is resolved in versions 1.2.17, 1.3.8, and 1.4.3.
Affected products
- Drupal AI (Artificial Intelligence) 0.0.0 to 1.2.17, 1.3.0 to 1.3.8, 1.4.0 to 1.4.3
Timeline
- 2026-06-24: advisory: Drupal security advisory SA-CONTRIB-2026-055 published
- 2026-07-10: disclosed: CVE-2026-13235 published to NVD