News·3 min read

GPT-5.6 Sol Ultra WordPress Pre-Auth RCE Using a Multi-Agent Exploit Chain

A critical pre-authentication remote code execution (RCE) vulnerability in WordPress has been uncovered not by a human researcher but by an AI system. The discovery, attributed to OpenAI’s GPT-5

CS
CyberShield Team
2026-07-20
Share:
GPT-5.6 Sol Ultra WordPress Pre-Auth RCE Using a Multi-Agent Exploit Chain

A critical pre-authentication remote code execution (RCE) vulnerability in WordPress has been uncovered not by a human researcher but by an AI system. The discovery, attributed to OpenAI’s GPT-5.6 Sol Ultra model, represents one of the most sophisticated automated vulnerability research, chaining a subtle validation bypass in WordPress’s Batch API into a full administrative takeover […] The post GPT-5.6 Sol Ultra WordPress Pre-Auth RCE Using a Multi-Agent Exploit Chain appeared first on Cyber Security News.

A critical pre-authentication remote code execution (RCE) vulnerability in WordPress has been uncovered not by a human researcher but by an AI system. The discovery, attributed to OpenAI’s GPT-5.6 Sol Ultra model, represents one of the most sophisticated automated vulnerability research, chaining a subtle validation bypass in WordPress’s Batch API into a full administrative takeover and code execution. The vulnerability originates in the WordPress Batch API (/wp-json/batch/v1), introduced in WordPress 5.6, which allows multiple REST requests to be bundled into a single call. GPT-5.6 Sol Ultra WordPress Pre-Auth RCE Rather than validating and executing each subrequest sequentially, the Batch API first runs validation across all requests, then executes them in a second pass. A flaw in this two-loop design means that when one request in the batch is malformed, the validation array and the “matched handler” array fall out of sync. This desynchronization lets an attacker validate one endpoint while executing a completely different one, effectively bypassing parameter sanitization on any batch-accessible endpoint. Sol identified that this validation mismatch could be paired with a pre-existing sanitization gap in the author_exclude parameter on GET /wp/v2/posts. When passed as a scalar string rather than an array, the parameter skips integer casting entirely and is interpolated directly into raw SQL. Since the Batch API doesn’t natively support GET requests, Sol engineered a recursive batch call nesting one batch request inside another to bypass that restriction and reach the vulnerable endpoint pre-authentication. Because leaked credentials in WordPress are hashed, database read access alone wasn’t enough for takeover. Sol instead exploited WordPress’s in-memory post cache, using UNION-based injection to fabricate fake post objects during a request. By abusing the oEmbed caching mechanism, which stores fetched content as oembed_cache posts and a post-hierarchy cycle-detection routine that resets parent references. Sol found a way to make WordPress reconcile a fabricated in-memory post with a real database row, effectively planting arbitrary post types and content into the database. This technique was used to forge a customize_changeset post, the mechanism WordPress uses to stage theme/settings changes tied to the administrator’s user ID. Applying the changeset temporarily elevated the request’s identity to admin. Sol then chained this into the parse_request action hook, which restarts the entire request-parsing pipeline mid-flight, allowing a pending “create administrator” request to succeed under the now-elevated privilege. With a rogue administrator account created, the final step was trivial: logging in and uploading a malicious plugin ZIP through the standard WordPress plugin installer, achieving code execution on the server. Searchlight cyber has published a free scanning tool at wp2shell.com to help site owners check exposure before public proof-of-concept exploits circulate more widely. Independent reproduction was reported by security researchers Calif and Hacktron shortly after disclosure. Given WordPress powers an estimated 500 million+ sites globally, and the flaw requires no authentication in a default MySQL-backed configuration, patching should be treated as urgent by any site administrator running affected versions. Prevent critical incidents and financial loss with stronger proactive defense. Integrate a live threat feed from 15K SOCs The post GPT-5.6 Sol Ultra WordPress Pre-Auth RCE Using a Multi-Agent Exploit Chain appeared first on Cyber Security News.

Share:

Join the Discussion

Comments coming soon. Follow us on social media for real-time discussions.