On August 30, Sybre Waaijer — the developer behind The SEO Framework — posted a thread on X that lit the WordPress ecosystem on fire. He revealed that Rank Math, one of the most popular SEO plugins running on over 4 million sites, had shipped an update that silently created WordPress Application Passwords with administrator-level permissions. No consent dialog. No notification. No opt-in checkbox. Just a quiet credential grab triggered the moment a user opened the plugin's Help & Support section while connected to a free Rank Math account.
The password — labeled "WAP – Rank Math Support Agent" in WordPress's Application Passwords panel — gave Rank Math's parent company group.one the ability to act on the site through their AI agent. It didn't expire. Closing the Help tab didn't revoke it. There was no toggle to disable it. If you opened that tab, you had it, whether you wanted it or not.
The same thing happened with WPForms a month earlier
This wasn't even the first time this pattern showed up in August. Waaijer had flagged a similar issue with WPForms Lite about three weeks prior. WPForms Lite, which runs on over 5 million sites, shipped a setup wizard that generated a one-hour login token and sent it to Awesome Motive's servers — again without asking. That token let their app install and activate plugins on the user's behalf, including third-party plugins from WordPress.org's repository.
The pushback on the WPForms allegations was split. Some users defended Awesome Motive as a trusted developer. Others pointed out that the principle matters regardless of who does it: a plugin should not create credentials and send them to external servers without explicit user consent. Period.
WordPress's own rules are clear about this
WordPress.org's plugin guidelines spell it out in section 7: plugins may not contact external servers without explicit and authorized consent, commonly done through an opt-in method. The Application Password feature in WordPress core is designed around a user-facing authorization screen — the plugin identifies itself, the user sees what's being requested, and they approve or reject it. Rank Math's implementation skipped all of that.
The transfer started before the Terms & Conditions box even appeared. The checkbox that was there didn't actually block the password creation. It was consent theater.
Rank Math's response
Within 48 hours, Rank Math paused the Support Agent. Their blog post acknowledged the failure directly: "When an AI agent is being given permission to create an application password, you should be told clearly and directly what is happening and what you're allowing."
They explained the intent — the Support Agent was meant to be read-only, answering questions with context from the user's actual site. It was supposed to be the first step toward a broader agentic AI deployment inside Rank Math, where AI agents could take actions on sites on behalf of users. The credentials were encrypted, not stored on their servers, and the agent inherited the current user's permissions rather than escalating them.
All of that might be technically accurate and still beside the point. The issue isn't what the credentials could do. It's that they were created without asking.
Rank Math confirmed the Support Agent is coming back with a redesigned consent flow that explains what's being requested in plain language before any credentials are generated. Whether users trust it the second time around is another question.
What you should do right now
If you use Rank Math and you've ever opened the Help & Support section while logged into a rankmath.com account, check your Application Passwords. Go to WP Admin → Users → Profile → Application Passwords and revoke anything starting with "WAP –". The password persists even after closing the support tab and has no expiration.
If you use WPForms Lite and went through the setup wizard recently, it's worth reviewing your audit logs and checking what plugins were installed during that session. The token was supposed to expire after an hour, but the principle of silent credential generation still applies.
More broadly: if you run WordPress sites for clients, this is a good moment to audit what your installed plugins are doing with external connections. The WordPress plugin ecosystem has long operated on an honor system where "trusted" developers are assumed to follow the guidelines. Two major plugins breaking the same rule in the same month suggests that assumption needs updating.
The bigger picture: AI agents need credentials, and that's the problem
What makes this story more than a one-week scandal is the pattern it reveals. AI agents are coming to WordPress — Rank Math said as much when they described the Support Agent as the first piece of a larger agentic AI strategy. AI agents need access to do useful work. Access means credentials. And credentials are exactly the kind of thing that should never be created silently.
This is the tension the entire AI tooling space is working through right now. The tools that win long-term are the ones where humans can inspect, approve, and revoke access at every step. The ones that skip that step — even with good intentions — erode the trust that makes the whole system work.
WordPress itself launched a new Core Security Initiative on September 1, partly in response to AI making it easier for attackers to find and exploit vulnerabilities. The irony of two major plugins creating undisclosed credential pathways in the same month the project announced a security push wasn't lost on anyone watching.
The lesson isn't that AI agents in WordPress are bad. It's that the credential layer around them needs to be explicit, inspectable, and revocable. Every time. No exceptions.
Sources: Rank Math Accused of Secretly Taking Admin Access — Search Engine Journal, Aug 30, 2026 · Rank Math Pauses Controversial Feature — Search Engine Journal, Aug 31, 2026 · WordPress Proactive Security Initiative — Search Engine Journal, Sep 1, 2026 · WPForms Lite Backdoor Allegations — Search Engine Journal, Aug 2026