Meta and Sierra are backing something called the Personal Agent Protocol (Poppy). It's supposed to create a standard for how personal AI agents interact with websites and businesses.

Sounds reasonable at first. But I was reading through the specification and some of the details seem pretty concerning from a privacy standpoint.

1. Websites can block AI agents they don't like

Section 4.1 allows companies to maintain lists of approved and blocked agents.

Sure there's a security argument for this. Nobody wants malicious bots accessing their systems.

But what happens when someone builds an independent AI agent that helps users find better deals or cancel subscriptions? A company could decide that agent isn't welcome.

It's a bit strange that something designed to represent the user could end up needing permission from the very company it's dealing with.

2.…