Executive brief
The Drupal AI Agents module, which allows websites to automate tasks like creating content or fields using AI, contains a security flaw. Under specific conditions, the module may incorrectly reuse parameters when an AI agent performs multiple actions in a single request. This could allow an unauthorized user to bypass access controls or view information they are not supposed to see.
Technical details
An incorrect authorization vulnerability (CWE-863) exists in the Drupal AI Agents module. The flaw occurs because the agent runtime may inherit deterministic parameters when invoking the same tool multiple times within a single request. This behavior can lead to forceful browsing and access bypass, potentially resulting in unauthorized information disclosure. The vulnerability is exploitable over the network, though the Drupal security team notes the attack complexity is high. Patches are available in versions 1.1.4, 1.2.5, and 1.3.1.
Affected products
- Drupal AI Agents 0.0.0 to 1.1.3, 1.2.0 to 1.2.4, 1.3.0
Timeline
- 2026-06-24: advisory: Drupal security advisory SA-CONTRIB-2026-057 published
- 2026-07-10: disclosed: CVE-2026-13237 published to NVD