Amazon’s decision to block Meta’s new Muse AI agent from shopping on its platform signals a much bigger battle over who controls customers, credentials and commerce in the agentic-AI era. Amazon says Meta did not notify it that Muse would access Amazon, the agent did not identify itself while browsing, and its handling of customer credentials raised security and privacy concerns. Meta, meanwhile, says Muse stores credentials securely and cannot see users’ passwords or payment details.
Key Highlights:
# Amazon blocks Muse over authorization, identification and security concerns
# User consent may not equal merchant consent
# AI agents threaten traditional e-commerce interfaces
# Agent identity and authentication could become mandatory
# Zero-Trust for AI agents could emerge as the next cybersecurity layer
The dispute exposes a fundamental question: does a customer’s permission automatically authorize an AI agent to transact with another company? Amazon says third-party agents should identify themselves and respect a service provider’s decision on whether to participate. Its 2026 agent policy similarly requires AI agents accessing Amazon services to clearly identify themselves.
The commercial implications are equally important. AI agents can potentially bypass product pages, recommendations, sponsored listings and other parts of the traditional shopping journey. Amazon generated more than $68 billion from advertising in 2025, giving it a strong incentive to retain control over product discovery and customer interaction.
Technologically, the confrontation could accelerate development of an identity layer for AI agents. Websites may increasingly demand machine-readable credentials identifying which agent is visiting, who authorized it, what permissions it has and whether it can browse, retrieve account information or execute transactions.
This could produce an “Agent Zero Trust” architecture: authenticate every agent, authorize every action, enforce least privilege, record transactions and require explicit approval for sensitive activities. Agent passports, signed requests, scoped tokens and auditable consent could become standard infrastructure.
The industry is therefore likely to move away from unrestricted browser automation toward permission-based agent commerce. Amazon’s earlier dispute with Perplexity shows the issue extends beyond Meta. Retailers, payment companies, browsers and AI developers will increasingly need interoperable rules governing agent identity, consent, data access and liability. The next internet battle may not be human versus AI—but platform versus agent.
