Drupal AI CVE-2026-13235

Exploitation requires the attacker to already hold a high-privileged authenticated session within the target Drupal installation (PR:H per CVSS vector), meaning an anonymous or low-privileged user cannot exploit this vulnerability. Attack complexity is rated high (AC:H), indicating that additional conditions beyond a simple authenticated request must be met - likely specific route configurations or module states. The Drupal AI (Artificial Intelligence) contributed module must be installed and enabled on the target site. The scope is unchanged (S:U), so compromise is limited to the Drupal application itself. Factors that limit exploitation: the attacker must obtain or already possess high-privilege credentials, the module must be present and active, and the attack cannot be automated at scale (SSVC automatable=no). No user interaction is required once those prerequisites are satisfied.

All available signals consistently place this vulnerability in the low-priority tier. The CVSS 3.1 base score is 3.3 (Low), driven by PR:H (high privileges required) and AC:H (high attack complexity), which substantially constrains the exploitable population to users who already hold significant trust within the Drupal installation. EPSS at 0.14% (4th percentile) corroborates that exploitation in the wild is essentially unobserved. CISA SSVC confirms exploitation=none, automatable=no, and technical impact=partial - reinforcing that mass or automated exploitation is not viable. There is no CISA KEV listing and no public POC. The limited C:L/I:L impact means even a successful exploitation yields only marginal confidentiality or integrity gains. The 'Authentication Bypass' tag in the intelligence data could be misleading: the CVSS PR:H metric confirms an attacker must already hold high privileges, so this is authorization bypass within an authenticated session, not a bypass of the authentication gate itself. The primary real-world risk is an insider or compromised high-privilege account accessing AI module resources beyond their intended scope - a scenario already partially mitigated by requiring that level of access in the first place.

An authenticated Drupal user holding a high-privileged role - such as an administrator or a role granted elevated AI module permissions - manually navigates directly to an AI module endpoint URL that lacks an authorization check, bypassing the intended access boundary and reading or modifying AI-related resources they were not explicitly granted access to. No exploit tooling or POC exists at time of analysis, meaning exploitation requires deliberate manual effort and prior knowledge of the target URL paths.