Amazon Tried to Block Perplexity's AI Shopping Bot — A Court Just Said No, and It Changes Everything
I've been watching the battle between Amazon and Perplexity unfold for months, and today it took a turn that I think has massive implications for the future of AI agents on the web.
The U.S. Ninth Circuit Court of Appeals just overturned an injunction that had banned Perplexity's Comet AI agent from shopping on Amazon.com. The ruling came down Tuesday, and the legal reasoning behind it is fascinating — and potentially a huge win for AI agent technology broadly.
How We Got Here
Here's the backstory: Back in late 2025, Amazon sent Perplexity a cease-and-desist letter demanding they block their AI browser, Comet, from accessing Amazon's store. Amazon claimed both companies had previously agreed to pause agentic shopping on the platform in 2024 — and that Perplexity broke that agreement by re-enabling Comet while disguising it as a regular Chrome browser.
Amazon filed suit in March 2026 and quickly won a temporary injunction banning Comet from Amazon. Perplexity appealed, and on Tuesday the Ninth Circuit overturned it.
The Legal Argument That Won
Here's where it gets interesting. Amazon argued Perplexity violated the Computer Fraud and Abuse Act (CFAA) — the federal computer-hacking statute — by having Comet "access" Amazon's computers without authorization.
The Ninth Circuit's response: Perplexity's tool didn't access Amazon's computers. The users did — with the help of the AI agent.
Because Comet only acts when a user tells it to, the court found it's the user who is doing the accessing. Perplexity is just the tool the user chose to use. Under CFAA, that distinction matters enormously.
This is a nuanced but important ruling. It essentially says that an AI agent acting on a user's instructions has the same authorization as the user themselves. The agent doesn't independently "access" anything — it's an extension of the user's intent.
What This Means for AI Agents
If this logic holds up going forward, it has sweeping implications for agentic AI technology:
- AI shopping agents can operate on e-commerce sites as long as users authorize them to.
- Websites can't use CFAA to block AI agents acting on behalf of authorized users.
- The key question becomes whether the user has permission, not whether the AI does.
This is a fundamental shift in how the law might treat AI agents. For years there's been uncertainty about whether an AI acting on your behalf has the same rights as you do on the web. The Ninth Circuit just said: yes, at least under CFAA, it does.
Amazon's Next Move
This ruling doesn't close the case. Amazon can request a rehearing from the full Ninth Circuit, or petition the Supreme Court to take the case. And the underlying lawsuit is still active — so even though the injunction is lifted, the legal fight continues.
Amazon clearly has strong incentives to fight this. If AI agents can freely shop on their platform on users' behalf, it disrupts one of Amazon's key competitive advantages: the friction that keeps users inside their ecosystem. When an AI agent can comparison-shop across multiple sites and complete purchases in seconds, Amazon loses some of the stickiness that makes its platform dominant.
Perplexity's Position
For Perplexity, this is a massive win — not just practically but strategically. Comet is central to Perplexity's vision of an AI browser that can act on the web on your behalf. Having a court validate that the CFAA doesn't block such agents from operating is huge for the entire category of agentic AI.
Perplexity CEO Aravind Srinivas has been positioning the company as building the next generation of internet search and browsing. A ruling that clears the path for AI agents to shop, research, and interact with web services on users' behalf is exactly the kind of legal foundation that category needs.
My Take
This case was always bigger than Amazon vs. Perplexity. It was really about whether the legal framework that governs computer access applies the same way when a human uses an AI to act for them.
The Ninth Circuit's answer — that the user is the one accessing, and the AI is just their tool — is the right one. It matches how we intuitively think about using software on our behalf. When you authorize your bank's app to access your accounts, the bank's servers aren't "accessing" anything. You are.
AI agents are the next evolution of that same principle. And now we have a federal appellate court saying so.
What's your experience? Drop a comment below! 👇 Do you use AI agents for shopping or web tasks? Do you think websites should be able to block AI agents acting on behalf of users?
Comments
Post a Comment