An authorization bypass vulnerability in the ash_ai library allows an authenticated caller to modify or destroy arbitrary records, posing a severe risk to data integrity.

This flaw is an Authorization Bypass Through User-Controlled Key (CWE-639) occurring in the AshAi.Tool.Execution module. By supplying crafted predicate expressions instead of literal values as tool arguments, an authenticated attacker can manipulate database queries to target records beyond their authorization scope.

The ability to arbitrarily update or delete database entries represents a critical risk to business operations and data reliability. A successful exploit could lead to unauthorized data destruction or modification, potentially resulting in significant operational downtime and loss of data integrity. With a CVSS score of 7.1, this vulnerability is classified as High severity, necessitating prompt attention to prevent unauthorized state changes in backend systems.

Immediate Action: Upgrade the ash_ai library to version 1.0.0 or later, which incorporates the necessary input validation to cast identity values to the correct field types.

Proactive Monitoring: Review application logs for unusual bulk update or destroy operations, specifically monitoring for tool execution patterns that include unexpected predicate structures in arguments.

Compensating Controls: Implement strict input validation or sanitization layers for all tool-configured inputs before they are passed to query execution functions to reject non-scalar inputs.

Public Exploit Available: Unknown

Given the potential for unauthorized data destruction, organizations utilizing the ash_ai library must prioritize updating to version 1.0.0. Security teams should verify their dependency trees to identify affected versions and ensure the patch is applied across all environments, as this vulnerability directly facilitates the manipulation of database records by authenticated callers.