A solo researcher chained a $500k-bounty-class WordPress RCE using GPT5.6 for $25

A solo security researcher just built a complete remote code execution exploit for WordPress using OpenAI’s GPT-5.6 for about $25. The exploit chain affects over 500 million WordPress sites and would fetch around $500,000 on the grey market. No major security outlet has covered it yet.

The attack strings together three vulnerabilities: a batch API desynchronization flaw, SQL injection, and automated admin account creation. It works pre-authentication, meaning an attacker needs no credentials to take over a site. The researcher used GPT-5.6 to identify the vulnerabilities, craft the exploit code, and automate the entire chain. Total API cost was roughly what you’d spend on half a week of ChatGPT Plus.

This matters for two reasons. First, if you run WordPress sites, you’re exposed until patches roll out and you apply them. That includes Enterprise DNA’s own web properties. The researcher disclosed responsibly, but the window between disclosure and patching is always risky. Second, the cost-to-impact ratio here is unsettling. A capable model, minimal spend, and one person produced an exploit that would previously require a team and weeks of manual work.

We’re not talking about theoretical risk. The original report shows working proof-of-concept code. The same capability that makes AI useful for building software now applies to breaking it, and the barrier to entry just dropped sharply. If you’re running any web infrastructure, this is the kind of threat surface you need visibility into. Tools like our Omni Command Centre can help monitor your stack and flag anomalies before they turn into breaches, but the broader point is simpler: patch fast, audit often, and assume attackers have access to the same models you do.