$500K WordPress RCEs: How Exploit Brokers Pay Big, and I Found One for $25 Using GPT-5.6

The cybersecurity landscape in mid-2026 has witnessed a startling revelation: exploit brokers are now paying up to $500,000 for a single WordPress Remote Code Execution (RCE) vulnerability. This news, detailed in a recent analysis by SLCyber, highlights the staggering economic incentives driving zero-day discovery. Even more striking, a researcher managed to identify one such critical flaw using only GPT-5.6 and $25 in API credits. This article dissects the mechanics of exploit brokerage, the technical nature of WordPress RCEs, and how AI tools are democratizing vulnerability research — while also raising urgent questions about security in the world's most popular CMS.

The Economics of Exploit Brokers

Exploit brokers operate as intermediaries between vulnerability researchers and buyers — often governments, law enforcement agencies, or private defense contractors. Unlike bug bounty programs (which pay for responsible disclosure), exploit brokers purchase vulnerabilities for weaponization and sale. According to the original SLCyber report, the top-tier price for a WordPress RCE has reached $500,000. This figure is not arbitrary: it reflects the difficulty of finding a reliable, unpatched RCE in a codebase scrutinized by thousands of developers and security researchers daily.

Source: SLCyber research center, July 2026.

The $500,000 price tag is reserved for vulnerabilities that require no authentication, work on default WordPress installations, and reliably execute arbitrary code on the server. These are the crown jewels for offensive security teams and surveillance vendors.

How GPT-5.6 Changed the Game

The SLCyber article describes a researcher who used GPT-5.6, the latest iteration of OpenAI’s language model, to accelerate vulnerability discovery. With a budget of just $25 in API credits, the researcher prompted the model to analyze WordPress core source code for known patterns of insecure deserialization and improper input sanitization. GPT-5.6, trained on a vast corpus of security research, CVE reports, and exploit code, generated a list of potential sink functions and tainted data flows. The researcher then manually validated the top candidate — a flaw in WordPress's REST API endpoint for media handling — and confirmed a working RCE.

Key capabilities of GPT-5.6 that aided this discovery:
- Static code analysis at scale: The model processed thousands of lines of PHP code in seconds, flagging dangerous functions like eval(), call_user_func(), and unserialize() with context awareness.
- Pattern recognition: GPT-5.6 identified a subtle bypass of a recent security patch by recognizing a logical error in input validation logic.
- Exploit scaffolding: The model generated a proof-of-concept PHP payload based on the discovered vulnerability, reducing development time from days to hours.

This workflow — AI-assisted code review followed by manual validation — is becoming a standard methodology for independent researchers. The $25 cost is a fraction of what a subscription to a commercial static analysis tool would cost, and the speed advantage is significant.

Technical Deep Dive: The WordPress RCE Found

While the exact CVE has not been publicly released (as the broker sale is ongoing), the article provides enough technical context to understand the vulnerability class. The discovered flaw resides in a WordPress REST API endpoint introduced in version 6.5. It involves improper handling of multipart form data when processing media uploads. Specifically, the wp_handle_upload() function fails to validate the MIME type of a file after it has been renamed, allowing an attacker to upload a .php file disguised as an image. The server then executes the uploaded script, giving the attacker full remote control.

This is a classic file upload RCE, but what made it worth $500,000 was the bypass of recent security hardening. WordPress 6.6 had introduced a check using finfo_file() to verify file signatures. However, the researcher found that by crafting a polyglot file — a valid JPEG with embedded PHP code — the signature check passed, while the file extension was manipulated via a double extension attack (e.g., shell.php.jpg). The server's .htaccess configuration did not block execution for such files.

The Role of AI in Modern Vulnerability Research

The use of GPT-5.6 for this discovery is not an isolated incident. Many researchers now integrate large language models into their toolchains for:
- Fuzzing seed generation: AI creates diverse, edge-case inputs for fuzzers like AFL++.
- Patch diffing: Models compare source code before and after a security update to pinpoint fixes — and potential bypasses.
- Exploit adaptation: When a PoC exists for an older version, GPT can adapt it to work against newer builds by analyzing changes in function signatures or data structures.

However, the article cautions that AI-generated code is not always correct or secure. The researcher manually verified every step of the exploit chain. GPT-5.6 hallucinated several false positives — suggesting vulnerabilities that did not exist under closer inspection. Human expertise remains irreplaceable for validation.

Implications for WordPress Site Owners

For the estimated 43% of websites running WordPress, this news is a wake-up call. The existence of a $500,000 RCE means that a critical, unpatched vulnerability is currently being traded in the underground market. Until the broker sells it to a buyer who may either use it or disclose it to WordPress (unlikely), millions of sites remain at risk.

Practical steps that administrators can take to reduce exposure:
1. Disable file editing in wp-config.php with define('DISALLOW_FILE_EDIT', true).
2. Restrict upload MIME types to only those needed, via upload_mimes filter.
3. Use a Web Application Firewall (WAF) with virtual patching for known REST API attack patterns.
4. Monitor for unusual file creation in wp-content/uploads/ — especially files with double extensions.
5. Keep WordPress core and plugins updated immediately upon patch release.

Ethical and Legal Considerations

The article also touches on the ethical dilemma for researchers. Selling a vulnerability to an exploit broker yields a high payout but contributes to the proliferation of zero-day weapons. Responsible disclosure to WordPress through its HackerOne bug bounty program would have earned the researcher a maximum of $10,000 — 50 times less. This disparity incentivizes the black market. The researcher in the SLCyber story chose the broker route, highlighting a systemic problem in vulnerability economics.

Conclusion

The revelation that exploit brokers pay $500,000 for WordPress RCEs — and that a single researcher found one using GPT-5.6 and $25 — underscores two converging trends: the skyrocketing value of web application zero-days, and the democratization of advanced vulnerability research through AI. As language models grow more capable, the barrier to entry for finding critical bugs will continue to fall. For defenders, this means the attack surface is expanding faster than ever. The only sustainable response is to adopt proactive security practices, including continuous code auditing, AI-powered defense tools, and community-driven patch management. The era of relying solely on reactive patching is over.

Source