ai identifies wordpress vulnerability

A frontier AI model has uncovered a critical WordPress vulnerability chain requiring less than $25 in compute to identify. GPT-5.6 Sol Ultra, directed at WordPress core source code by security researcher Adam Kues from Assetnote at Searchlight Cyber, surfaced a pre-authentication remote code execution path now labeled wp2shell. The discovery prompted emergency patches and raised immediate questions about AI-assisted vulnerability research economics.

The wp2shell chain combines two distinct flaws. The first is a REST API batch endpoint confusion bug (CVE-2026-63030) that allows crafted requests to bypass standard input validation. The second is a SQL injection vulnerability in a core query parameter (CVE-2026-60137), previously reported by researchers TF1T, dtro, and haongo. Together, the flaws allow an unauthenticated attacker to sequence requests that achieve unauthorized database access, then escalate through a code gadget chain to administrator-level control and arbitrary code execution.

WordPress classifies the chain as one critical and one high-severity vulnerability.

GPT-5.6 identified the SQL injection component as its primary high-value finding during the audit. Kues then spent approximately four hours converting that AI-identified flaw into a fully functional remote code execution exploit chain. Total compute expenditure for the AI-assisted discovery phase reached roughly $25, a threshold that considerably lowers the barrier for high-impact vulnerability research compared to traditional manual auditing timelines.

Affected releases span the 6.9 and 7.0 WordPress branches. Versions 6.9.0 through 6.9.4 and 7.0.0 through 7.0.1 are vulnerable to the complete unauthenticated remote code execution chain on default installations. The 6.8 branch carries exposure only to the underlying SQL injection component and lacks the conditions necessary for full exploitation.

WordPress versions predating 6.8 are reported as unaffected. Risk is concentrated on sites running current default configurations without additional hardening or compensating controls, representing a substantial portion of the global WordPress install base.

Public proof-of-concept code has been reported for the affected releases, increasing urgency for site operators to apply patches. WordPress shipped emergency security updates on July 17, 2026, addressing both CVEs across the impacted branches. Site administrators running any version within the 6.9.0–6.9.4 or 7.0.0–7.0.1 ranges on default configurations are advised to update immediately.

The wp2shell case illustrates a broader shift in exploit economics. An AI model consuming $25 in compute, followed by several hours of researcher effort, produced a validated critical vulnerability chain in widely deployed software. Professional plugin security audits covering up to 10,000 lines of code are available for $250 full review, offering site owners a structured alternative for identifying similar risks in custom or third-party code before attackers do.

That combination compresses what traditionally required extended manual code review into a narrow and inexpensive workflow. Security teams and software vendors now face an environment where AI-assisted discovery can surface severe vulnerabilities faster and at lower cost than conventional research cycles, demanding faster patch development and deployment responses in turn.

You May Also Like

Google-Backed FireSat Satellites Use AI to Detect Wildfires Faster

Google-backed FireSat satellites use AI to detect wildfires as small as 5×5 meters, and what they’ve already found may surprise you.

OpenAI Launches GPT-Red: How the New AI “Super-Hacker” Could Strengthen Cybersecurity

The AI “super-hacker” GPT-Red is reshaping cybersecurity with an 84% attack success rate—but its most surprising impact may be yet to come.

Capital One Releases VulnHunter, an Open-Source AI Tool for Detecting Software Vulnerabilities

Discover how Capital One’s open-source AI tool, VulnHunter, is transforming software security by exposing vulnerabilities faster than ever before.

AI Spam Filters Remain Vulnerable to Old-School Text Salting Attacks

Invisible text tricks are outsmarting today’s most advanced AI spam filters, and the implications for email security are more alarming than you’d expect.