WordPress shipped another security release, and attacks on the new PHP flaws kept growing. The WordPress AI team also published a plan to let AI agents connect to any site by 7.3.

Security: 7.1.4 and the PHP Follow-Up

WordPress 7.1.4 arrived on October 28. It fixes a weakness in the upgrade process that can lead to remote code execution, meaning an attacker runs their own code on your server. Umut Bayram reported it. The release also fixes several stored cross-site scripting flaws, which let malicious script saved on your site run in a visitor’s browser.

Sites with automatic background updates will already have it. Everyone else should open Dashboard, then Updates, and update now. WordPress.org is backporting the fixes to older branches, going back to 4.7.

This is the seventh core security release since July. Our coverage shows the run: 7.0.2, 7.0.3, 7.0.4, 7.1.1, 7.1.2, 7.1.3, and now 7.1.4. The cause is a flood of bug reports, many found with AI tools, which we covered in August. Expect updates every two to three weeks, and keep auto-updates on.

The PHP flaws from last week are still being probed. Wordfence says it has blocked more than 74,000 attempts and has seen no confirmed break-ins. Some attackers send crafted upload requests that try to read wp-config.php and .env, the files that hold your database password and secret keys.

WordPress updates do not fix PHP. Make sure your host installed PHP 8.5.12, 8.4.27, 8.3.36, or 8.2.35, and restarted its PHP workers. An update that is installed but not restarted may not protect you.

AI: A Four-Part Plan for Agent-Ready WordPress

On October 28, Automattic’s AI team representative Jason Adams laid out a plan to make WordPress ready for the AI agents people bring with them. The target is WordPress 7.3, with groundwork during the current 7.2 cycle.

His example: you tell your own AI assistant to help with your site. The assistant spots WordPress and sends you a link. You sign in, approve access, and it gets to work.

The plan has four parts:

  • MCP: the standard that lets AI assistants use a site’s features. The team may merge the adapter into core or bundle it.
  • Abilities: the permission-checked actions an agent can run. The team is defining a baseline for common site tasks.
  • OAuth: a sign-in flow that replaces Application Passwords, the manual credentials users create today.
  • User experience: the full path, from an agent finding a site to you revoking its access.

This builds on the news from last week’s issue that Matt Mullenweg approved building MCP into core. The MCP Adapter reached the plugin directory earlier this month. Adams wrote that it asks too much of users, who must install it, find useful Abilities, and manage Application Passwords. The AI team says it needs contributors to lead the work.

Nothing here changes your site today. If you plan to let an AI agent in, the sign-in approach should be safer than sharing a long-lived password. Watch the 7.2 betas for early pieces.

End of article