WordPress Plugin Updates Now Get AI Security Reviews

WordPress Adds an Automated Checkpoint for Plugin Updates

Every plugin release published through WordPress.org now passes an automated security review before it can reach websites. The WordPress plugin repository team announced the change on September 9: during the existing cooldown window, multiple AI models and Automattic's Jetpack Scan analyze each release's code changes, and high-risk releases are blocked automatically. Given how much of the web runs on WordPress plugins, it is one of the widest-reaching supply chain defenses to ship this year.

• Coverage: every plugin release before it is distributed through the WordPress.org update API
• Timing: analysis runs during the release cooldown, currently 6 hours, which was introduced on June 5, 2026
• Method: several AI models plus Jetpack Scan review the code changes, and their results are cross-checked and combined into a security score
• Outcome: high-risk releases are blocked automatically, and the plugin's committers get an email with the findings

Why Review Every Release, Not Just New Plugins?

New plugins have long gone through human review before they are listed. The harder problem is what happens afterward. As David Perez, co-lead of the WordPress plugin repository team, puts it in the announcement, a plugin can be secure today and introduce a vulnerability, or malicious code, in a future release. A compromised developer account or a careless commit can push a problem to every site that auto-updates.

The cooldown already proved its value. On July 28, a backdoor was found in a plugin with roughly 20,000 installations. Because the compromised release was still sitting in the cooldown window, it never went out through the update API, and Help Net Security reports the plugin was closed 26 minutes after Wordfence flagged it. Automated review turns that holding period from a waiting room into an active inspection.

How Does the AI Security Review Work?

The design is layered, which is the right call for automated security tooling. No single AI model gets the final word. Multiple models and a conventional scanner evaluate the same diff, and cross-checking their results is meant to catch real problems while keeping false positives down. The combined output becomes a risk score, and only releases that score high are held back.

Importantly, a high score is not an accusation. The announcement stresses that a vulnerability can score just as high as intentional malware, so a blocked release says the code needs another look, not that its author did anything wrong.

What Happens When a Plugin Update Gets Blocked?

The release stays out of the update API, so sites never receive it. The committers get an email describing what was found. From there they can review the findings, fix the issue and publish a new version, which goes through the normal cooldown and review again. Developers who believe a finding is mistaken can contact the Plugins Team, though the team notes that shipping a corrected release is almost always faster than an appeal.

For site owners, nothing changes except that updates are safer. For plugin developers, it is worth building a pre-release scan into your own pipeline so the WordPress.org check becomes a formality rather than a surprise. The team says it will keep tuning the system with real data, and it is asking developers to report false positives so the scoring improves.

Signing and verifying what ships is a theme across open source right now, from Chainloop signing every artifact a CI pipeline builds to the new ORKS standard for revocable API keys. Follow more defensive tooling in our AI security coverage.

Sources: Make WordPress Plugins — September 9, 2026; Help Net Security — September 10, 2026.