CVE-2025-13922 Overview
CVE-2025-13922 is a time-based blind SQL injection vulnerability in the Tag, Category, and Taxonomy Manager – AI Autotagger with OpenAI plugin for WordPress. The flaw affects all versions up to and including 3.40.1. The vulnerability resides in the AI preview AJAX endpoint, where the existing_terms_orderby parameter is passed into a SQL query without proper escaping or parameterization [CWE-89]. Authenticated attackers with Contributor-level access or higher who hold AI metabox permissions can append arbitrary SQL to the executing query. Successful exploitation enables extraction of sensitive database contents, performance degradation, and data inference through timing side channels.
Critical Impact
Authenticated attackers with Contributor-level privileges can exfiltrate confidential database content through time-based blind SQL injection against the AI preview AJAX endpoint.
Affected Products
- Tag, Category, and Taxonomy Manager – AI Autotagger with OpenAI (simple-tags) plugin for WordPress
- All plugin versions up to and including 3.40.1
- WordPress sites allowing Contributor-level or higher accounts with AI metabox permissions
Discovery Timeline
- 2025-12-06 - CVE-2025-13922 published to NVD
- 2026-10-15 - Last updated in NVD database
Technical Details for CVE-2025-13922
Vulnerability Analysis
The vulnerability exists in the TaxoPress AI AJAX handler responsible for previewing AI-suggested taxonomy terms. When processing the AI preview request, the plugin reads the existing_terms_orderby parameter from the client and injects it directly into a SQL ORDER BY clause. Because ORDER BY positions cannot be bound via prepared statement placeholders, developers must apply strict allowlist validation. The plugin fails this requirement, instead concatenating the attacker-controlled value into the final query string.
An authenticated attacker invokes the AJAX endpoint and supplies a crafted existing_terms_orderby value containing stacked SQL expressions. Blind extraction is achieved using conditional time delays, such as SLEEP() or heavy subqueries, enabling the attacker to infer database contents one bit at a time. Impact is confined to confidentiality under the published scoring, but the exposed data includes WordPress user records, session tokens stored in wp_usermeta, and plugin-managed secrets.
Root Cause
The root cause is insufficient escaping of a user-supplied ORDER BY parameter combined with the absence of SQL query parameterization. The vulnerable code path exists in inc/class.admin.php around line 1406 and in modules/taxopress-ai/classes/TaxoPressAiAjax.php around line 180, as referenced in the WordPress Plugin Code Review and the WordPress Plugin Ajax Class.
Attack Vector
Exploitation requires network access to the WordPress site and an authenticated session with Contributor privileges or higher, plus the AI metabox capability. The attacker issues an admin-ajax.php request to the plugin's AI preview action and supplies a malicious existing_terms_orderby value. No user interaction from administrators is needed. See the CleanTalk CVE Analysis and Wordfence Vulnerability Report for additional technical context.
Detection Methods for CVE-2025-13922
Indicators of Compromise
- POST requests to /wp-admin/admin-ajax.php carrying the TaxoPress AI preview action and anomalously long existing_terms_orderby values
- Request bodies containing SQL keywords such as SLEEP, BENCHMARK, UNION, SELECT, or IF( within the existing_terms_orderby parameter
- Repeated AJAX requests from a single Contributor account producing unusually high backend latency
- Spikes in MySQL query execution time correlated with authenticated plugin AJAX traffic
Detection Strategies
- Deploy web application firewall rules that inspect the existing_terms_orderby parameter for SQL metacharacters and function names
- Correlate WordPress audit logs with database slow-query logs to flag Contributor accounts triggering long-running queries
- Alert on sequential AJAX requests to the TaxoPress AI endpoint exhibiting binary-search timing patterns characteristic of blind SQL injection
Monitoring Recommendations
- Enable MySQL slow query logging with a threshold aligned to normal application latency
- Monitor WordPress plugin activity and privilege changes for Contributor and Author accounts
- Track outbound data volume from the database host to detect staged exfiltration following inference-based extraction
How to Mitigate CVE-2025-13922
Immediate Actions Required
- Update the Tag, Category, and Taxonomy Manager – AI Autotagger plugin to the version released after 3.40.1 as published in the vendor WordPress Plugin Changeset Log
- Audit all Contributor, Author, and Editor accounts and remove any that are inactive or unexpected
- Rotate WordPress secret keys, administrator passwords, and any API tokens stored in the database
Patch Information
The vendor addressed the vulnerability by introducing allowlist validation for the existing_terms_orderby parameter in the TaxoPress AI AJAX handler. Review the fix via the WordPress Plugin Changeset Log and confirm the installed version exceeds 3.40.1.
Workarounds
- Temporarily deactivate the plugin until patched installations can be verified across all sites
- Restrict AI metabox permissions to trusted administrator roles only
- Deploy virtual patching through a WAF to block requests containing SQL functions in the existing_terms_orderby parameter
- Enforce strong authentication, including multi-factor authentication, for all Contributor and higher accounts
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
