Overview
- Amazon began blocking Muse from performing shopping actions on its site and showing users a popup that the AI’s access is unauthorized after traffic from the agent tried to place orders on the platform on Monday.
- Amazon says Meta never obtained permission, that Muse does not identify itself during browsing, and that the agent appears to collect or store user account credentials, which the company says violates its terms of service and raises security and commercial concerns.
- Meta counters that Muse runs each user in an isolated virtual machine, stores login data in a secure vault so the agent cannot read plaintext passwords or payment data, and uses an independent ‘Sentinel’ monitor to approve external requests and sensitive operations.
- Shopify and the Universal Commerce Protocol are moving in the opposite direction by endorsing agent checkout flows and listing Muse as supported, creating an alternate, interoperable path for agent‑led purchases with payments partners like Stripe, Visa and Mastercard.
- A recent Ninth Circuit ruling limits criminal hacking claims against agent makers, leaving platform terms and commercial arrangements as the main levers, which means the dispute is likely to be settled by platform enforcement, direct talks between Amazon and Meta, or wider industry alignment that will shape how consumers trust and use shopping agents.