Amazon blocks Meta's Muse — and it's fencing out shopping agents
The block cites unauthorized-agent ToS; Amazon already fended off Perplexity, Google and OpenAI. Building browsing agents? Site access is now a dependency.
Copy markdown
What Amazon actually did
Since Sunday night, shopping on Amazon through Meta's Muse returns a popup: “Continued access by an unauthorized AI agent violates Amazon's Conditions of Use.” Muse launched earlier this month and drives real sites through a browser the way a person would; Amazon says Meta never disclosed the access and the agent doesn't identify itself when it browses.
The real reasons: ad money and liability
Amazon booked $68B in ad revenue last year off shoppers seeing sponsored products — an agent that just buys skips all of it. It also flags security: Muse can reach account pages and order history on request, an “undisclosed third party moving through customer accounts.” Meta counters that Muse runs in a secure VM with a “Sentinel” monitor and no visibility into passwords or payment methods.
A pattern, not a one-off
Amazon has already sued Perplexity over its Comet browser and blocked shopping agents from Google and OpenAI. Its own “Buy for Me” agent, by contrast, identifies itself and lets brands opt out. The signal to builders: first-party, identified agents are welcome; unidentified third-party ones get fenced out.
The missing piece: agent authentication
There's still no standard for an agent to identify itself or negotiate authorized access, so browser-mimicking agents look identical to humans — until a platform fences them off by Terms of Service rather than anti-hacking law (the Ninth Circuit put site access on the user, not the AI company). If you're building browsing or checkout agents, design for explicit, per-site permission now.