Critical Vulnerabilities in WordPress Exploited Using Advanced AI Technology
Researchers at Searchlight Cyber have harnessed the capabilities of OpenAI’s GPT5.6 Sol Ultra to successfully devise a comprehensive exploit chain targeting two significant vulnerabilities within the WordPress Core framework.
The first vulnerability, identified as CVE-2026-63030, is a severe issue related to REST API batch endpoint route confusion, earning a disconcerting CVSS rating of 9.8. This flaw affects WordPress Core versions 6.9.x prior to 6.9.5 and 7.0.x before 7.0.2.
The second vulnerability, marked as CVE-2026-60137, pertains to a high-severity author__not_in WP_Query SQL injection, with a CVSS rating of 5.9. It impacts WordPress Core versions 6.8.x before 6.8.6, as well as 6.9.x before 6.9.5 and 7.0.x before 7.0.2.
On July 17, WordPress released version 7.0.2 of their content management system (CMS) to address these critical vulnerabilities.
By chaining the exploits for both vulnerabilities, attackers can execute remote code pre-authentication on WordPress installations running the vulnerable versions 6.9.x and 7.0.x.
Searchlight Cyber, credited with the discovery of CVE-2026-63030, coined the exploit chain ‘WP2Shell’ in an advisory released on July 17.
Unlike the majority of WordPress vulnerabilities, WP2Shell affects the core framework directly rather than third-party plugins or themes. Consequently, unauthenticated attackers can exploit these vulnerabilities on a default WordPress installation.
“This attack has no prerequisites, making it vulnerable to anonymous users utilizing a standard WordPress setup devoid of plugins,” commented Adam Kues, a security researcher at Searchlight Cyber.
GPT-5.6 Sol Ultra Discovers Elaborate Pre-Authentication RCE Path
Kues published a detailed report on July 20, outlining how the GPT5.6 Sol Ultra model was employed to construct the exploit.
To unravel the exploit chain, Kues repurposed a prompt used by OpenAI for solving intricate mathematical problems.
He meticulously removed the version history from a pristine copy of the WordPress codebase to thwart the model from providing compromised outputs.
He then instructed it to deploy up to four agents running for a minimum of six hours to search for pathways leading to pre-authentication remote code execution.
The model promptly identified the REST API batch route confusion (CVE-2026-63030), understanding that a disjunction between validation and execution enabled it to circumvent input sanitization, thereby triggering the SQL injection (CVE-2026-60137).
In mere minutes, the model leveraged the SQL injection to extract the administrator’s email from a fresh installation.
However, the model’s true prowess manifested in its ability to escalate this read-only database access to complete remote code execution.
Over an additional four hours, the AI engineered a remarkably intricate, multi-layered exploit chain. It utilized cache poisoning via counterfeit oEmbed entries, manipulated WordPress customize changesets to temporarily acquire administrative privileges, and invoked hooks that permitted it to upload a backdoor plugin without authentication.
This fully automated process required just over ten hours and incurred approximately $25 in computational costs, a mere fraction of a typical premium subscription, according to Kues.
The resulting exploit chain was astoundingly complicated, likely requiring weeks, if not months, for a human researcher to assemble independently.
Kues remarked, “No security researcher could replicate this exploit chain within 10 hours without the assistance of AI.”
With exploit brokers reportedly willing to pay up to $500,000 for a zero-day remote code execution vulnerability in WordPress Core, Kues emphasized that the discovery of such a flaw through an off-the-shelf AI model underscores a significant shift in the economic landscape of cybersecurity.
Following the public disclosure of the two WordPress Core vulnerabilities, several other security analysts have begun releasing proof-of-concept (PoC) exploits.
Additionally, on July 17, security firm PatchStack reported exploitation of both CVEs; however, the specific activities observed in the wild remain unclear.
Other organizations, including Hexastrike and WatchTowr, have reported signs of exploitation attempts occurring in the wild.
WordPress Implements Automatic Updates to Counter WP2Shell
In an extraordinary measure, WordPress has mandated automatic updates for affected installations to swiftly address the threat.
Security administrators are urged to manually confirm that their sites are operating on version 7.0.2 or version 6.9.5.
Furthermore, Searchlight Cyber has introduced a complimentary scanning tool available at wp2shell.com, enabling administrators to safely assess their servers for vulnerability against this AI-engineered attack.
Source link: Infosecurity-magazine.com.
