On October 6, 2026, WordPress pushed version 7.1.3 as a security and maintenance release. Seven vulnerabilities patched, four bugs fixed, and an urgent recommendation to update immediately. Standard stuff for a security release — except for one detail that changes how you should think about who is looking at your site's code. Three of the seven vulnerabilities were reported by AI companies. Not by independent security researchers, not by bug bounty hunters, by OpenAI and Anthropic.
That is worth pausing on. The companies building the AI tools that increasingly interact with your WordPress site are also the ones auditing its security. They are not doing it out of charity. They are doing it because their models need to work with WordPress, and they need that code to not be full of holes.
What got patched
The seven vulnerabilities in WordPress 7.1.3 range from annoying to genuinely dangerous. The most concerning is a stored XSS on the Comments administration page, exploitable through pending comments. Stored XSS means an attacker can inject malicious code that gets saved to your database and then executes every time an admin views the comments section. Trail of Bits found this one in collaboration with OpenAI. If you accept comments on your site and an admin ever looks at them — which is the whole point of comments — this vulnerability had a direct path to execution.
The second critical finding is a second-order SQL injection in WordPress's WXR export feature. Second-order SQL injection is nastier than the regular kind because the malicious input does not execute immediately. It gets stored first, then triggers when a different process reads and uses that data. In this case, an attacker could inject SQL through content that would execute when someone exported the site's data. Anthropic reported this one.
Anthropic also found a DoS vulnerability in the WP_Http::make_absolute_url() method and a weakness that lets Author-role users sticky posts they should not be able to sticky. The DoS issue could theoretically let an attacker crash or hang a WordPress process by feeding it a crafted URL. The sticky posts weakness is lower severity but still a real privilege escalation — an Author should not be able to pin their own content to the top of your homepage.
The remaining three vulnerabilities were reported by independent researchers. Ananda Dhakal from Patchstack found an unauthenticated disclosure issue that could leak comments on private and unpublished posts. Three researchers identified XSS in Imgur embed handling. And Alex Concha from the WordPress security team found that forgeable parameters could cause action name collisions in WordPress hooks.
Why AI companies are doing security research now
This is the first WordPress security release where OpenAI and Anthropic show up as vulnerability reporters, and it is probably not the last. The reason is practical. Both companies are building AI products that interact with WordPress. OpenAI's GPT-6 launched on October 7 with capabilities that include interacting with web applications. Anthropic's Claude is a core part of the MCP ecosystem that WordPress just officially adopted. When your product depends on a platform being secure, you have a direct financial incentive to find the cracks before someone else does.
This is also a new chapter in how open-source security works. Traditionally, vulnerabilities get found by security firms, independent researchers, or the occasional developer who stumbles into something while debugging. AI companies entering the field as active auditors changes the economics. They have the resources to run automated security analysis at scale. They have models that can reason about code paths and identify edge cases that static analysis tools miss. And they have a product reason to keep going.
The WordPress security team is clearly happy about this. The official announcement lists OpenAI and Anthropic alongside Trail of Bits and Patchstack without any special distinction. They are just more researchers doing responsible disclosure. But the scale is different. OpenAI and Anthropic are not individuals filing reports in their spare time. They are organizations with dedicated security research capacity that happens to intersect with the platforms their AI products need to work with.
The practical checklist for your site
If you are running WordPress, the action item is straightforward: update to 7.1.3 now. Not later today, not this weekend, now. The stored XSS alone is enough to justify immediate action. If you accept comments and any admin on your site views pending comments, that vulnerability has a clear exploitation path.
Here is how to handle the update depending on your setup.
If your site supports automatic background updates, the process may already be running. WordPress 7.1.3 is a point release, which means it qualifies for automatic updates by default. Check your dashboard under Updates to confirm. If you see 7.1.3 listed as the current version, you are already patched.
If you manage updates manually, go to Dashboard, then Updates, then click Update Now. If you are running a caching plugin, clear your cache after the update to make sure visitors are not served stale pages from before the patch.
If you manage multiple WordPress sites, check whether your hosting provider has a bulk update tool. Many managed hosts push security updates automatically, but not all of them do. Do not assume. Verify each site individually.
One more thing: WordPress is backporting these security fixes to older branches going back to version 4.7. If you are running an older WordPress installation that you have not updated in a while, the patches are coming. But "coming" is not the same as "arrived." If your site is on an older branch and the backport has not landed yet, the safest move is to update to 7.1.3 if your theme and plugins support it.
The bigger shift this signals
WordPress 7.1.3 is a routine security release. The vulnerabilities are serious but not unusual in type or severity. What is unusual is who found them and why. AI companies are now actively auditing the platforms their models interact with, and that is a structural change in how open-source security will work going forward.
For site owners, this is mostly good news. More eyes on the code means more vulnerabilities get found and fixed before attackers exploit them. But it also means the pace of security releases may increase as these new auditors find more issues. The days of ignoring WordPress updates for months at a stretch are over. If AI companies are running continuous security analysis on the codebase, the vulnerability discovery rate is going to go up, and so is the frequency of patches you need to apply.
The smart move is to enable automatic updates if you have not already and make sure your hosting environment supports them. The era of "I will get to it eventually" for WordPress updates ended on October 6.
Sources: WordPress.org — WordPress 7.1.3 Maintenance and Security Release (October 6, 2026). Search Engine Journal — WordPress 7.1.3 Fixes 7 Vulnerabilities And 1 Critical Flaw (October 7, 2026).