The official AI plugin 1.5.0 adds six experiments and two security fixes. If you run the plugin, update it. Then check which experiments are switched on.
The six new experiments
- Text to Speech. Generates an audio version of a post using an AI provider that supports speech. It runs in the background and saves the file to your Media Library. You choose whether visitors see a player.
- Content Verification. Detects provenance labels called C2PA Content Credentials in uploaded JPEG, PNG, and WebP images. It only reads them and does not verify or create them.
- MCP Access. Connects external AI agents to your site through the Model Context Protocol. It adds a Tools, MCP Access screen where you pick which abilities agents may use. It can install the MCP Adapter plugin if permissions allow.
- WebMCP. Lets compatible AI browsers read and edit the post title and blocks, and save or publish. It is experimental.
- Revision Change Summary. Writes a plain-language summary of the differences between two revisions.
- Block Container Label Generation. Suggests short labels for Group and Columns blocks in List View.
Other changes and what to do
Mistral and Vercel AI Gateway join the list of AI providers. The settings now separate stable features from experiments and flag those that run automatically.
Markdown feeds changed. URLs ending in /feed/markdown/ no longer work. Use ?output_format=markdown instead. Three older ability names were also removed, which matters if you use custom integrations.
The release includes fixes for two security advisories. Update first, then review the experiments and keep MCP Access limited to what you need.
End of article