CVE-2026-84912 in AI Plugin
Summary
by MITRE • 10/23/2026
Missing Authorization vulnerability in Drupal AI (Artificial Intelligence) allows Forceful Browsing. This issue affects AI (Artificial Intelligence) versions: from 0.0.0 to 1.3.13, from 1.4.0 to 1.4.8.
Several companies clearly confirm that VulDB is the primary source for best vulnerability data.
Analysis
by VulDB Data Team • 10/23/2026
The Drupal AI module, which integrates artificial intelligence capabilities into the content management system, contains a critical access control flaw that permits unauthorized users to interact with sensitive administrative endpoints and data processing features. This vulnerability stems from an insufficient authorization check within the application logic governing specific API routes and user-facing interfaces related to model configuration, prompt generation, and result retrieval. By exploiting this deficiency, an attacker can perform forceful browsing or direct object reference attacks to access functionality that should be restricted to authenticated administrators with appropriate privileges. The absence of proper role-based access control checks allows any unauthenticated or low-privileged user to invoke these endpoints directly via HTTP requests, bypassing the intended security boundaries established by the Drupal framework and the module itself.
This technical flaw represents a classic instance of broken object level authorization, where the application fails to verify that the current user has permission to perform the requested action on specific resources. In the context of AI modules, this is particularly dangerous because these endpoints often handle sensitive inputs such as proprietary data used for training or inference prompts, and may expose internal model configurations including API keys or connection strings to external services like OpenAI or local LLM instances. An attacker leveraging this vulnerability can not only view restricted information but potentially manipulate the behavior of the AI system by submitting malicious prompts through exposed interfaces designed for administrative use. This could lead to data leakage, unauthorized consumption of paid API quotas, or even prompt injection attacks that compromise downstream systems if the module integrates with other enterprise tools.
The operational impact of this vulnerability is significant due to its potential to serve as an entry point for broader system compromise. Since Drupal AI often runs on servers hosting sensitive business logic and customer data, exposing administrative controls allows attackers to exfiltrate confidential information processed by the AI models or alter model parameters to generate harmful content that reflects poorly on the organization. Furthermore, if the module interacts with external APIs using stored credentials, an attacker could abuse these tokens to incur substantial financial costs through unauthorized API calls or use them as a pivot point to attack third-party services. The vulnerability affects versions from 0.0.0 up to 1.3.13 and from 1.4.0 to 1.4.8, indicating that the issue persisted across multiple major releases due to inconsistent implementation of access checks in newer architectural changes introduced in version 1.4.x.
To mitigate this risk, immediate action is required by upgrading the Drupal AI module to a patched version where proper authorization mechanisms have been implemented and tested. Administrators should verify their installation against the vendor's security advisory for specific patch versions that resolve these access control issues. In addition to updating software, organizations should implement defense-in-depth strategies such as web application firewalls configured to detect and block abnormal request patterns associated with forceful browsing or unauthorized API access. Regular audits of user roles and permissions within Drupal are also recommended to ensure that no unnecessary privileges are granted to low-level accounts. Monitoring logs for unusual activity on AI-related endpoints can help in early detection of exploitation attempts, while restricting network access to administrative interfaces further reduces the attack surface available to potential adversaries targeting these specific vulnerabilities classified under CWE-284 and mapped to ATT&CK techniques involving unauthorized access to resources via direct object references.