The future of AI shopping just ran into a major obstacle.

Amazon has blocked Meta's Muse AI agent from shopping on its platform, citing concerns around authorization, security, privacy and transparency.

The dispute highlights a fundamental problem with agentic AI: Who controls the interaction when an AI acts on behalf of a customer?

Why did Amazon block Muse?

Amazon says Muse was accessing its website without authorization and that the agent did not clearly identify itself while browsing.

Amazon also raised concerns about credentials and account data.

Meta, meanwhile, says Muse is designed with security controls and that the agent does not have visibility into passwords or payment methods.

The disagreement exposes a new category of platform conflict.

Traditional websites were built for human users and conventional software integrations. AI agents introduce a third type of visitor: software that behaves like a user and can make decisions while navigating websites.

Why this matters beyond Amazon

If AI agents become common shopping assistants, websites will need to decide how these agents should interact with their platforms.

Should agents be allowed?

Should they require special APIs?

Should websites charge for automated access?

How should authentication work?

And who is responsible if an agent makes an incorrect purchase?

These questions have not been fully standardized.

The bigger lesson

Agentic AI needs more than model intelligence.

It requires identity, authentication, permission management, auditability and clear rules between agents and platforms.

The Amazon-Muse dispute is therefore bigger than one shopping feature.

It could become an early example of how the internet needs to adapt when software agents become active participants rather than passive tools.