Amazon Blocks Perplexity AI Shopping Agent in Court

The Perplexity Comet browser just got hit with a federal injunction, and the ruling cuts right to the heart of who actually controls what AI agents can do online. U.S. District Judge Maxine Chesney sided with Amazon on Monday, blocking Comet from executing purchases inside Amazon accounts on users' behalf — at least until a full trial can determine whether the Computer Fraud and Abuse Act even applies to agentic software acting on a human's explicit instruction.

What Did the Court Actually Rule?

Why Amazon won the preliminary injunction against Comet

Judge Chesney found that Amazon had provided what she called "essentially undisputed evidence" that Perplexity accessed password-protected Prime accounts — with users' knowledge, yes, but without Amazon's authorization. That gap between user permission and platform permission is the whole ballgame here. Perplexity's defense was straightforward: Comet just does what users tell it to do, so it inherits their access rights. The court, at least at this stage, didn't buy it.

Under the injunction, Perplexity Comet browser must immediately stop accessing those accounts and destroy any Amazon customer data already harvested through the tool. Perplexity has seven days to file an appeal with the Ninth Circuit before the order bites.

We will continue to fight for the right of internet users to choose whatever AI they want.

How Did the Amazon Perplexity Lawsuit Start?

The Amazon Perplexity lawsuit was filed in November 2025, but the hostilities started earlier. Amazon says it warned Perplexity at least five times beginning in November 2024 to stop Comet from disguising its automated traffic as regular Google Chrome browser sessions. When Amazon finally deployed a technical block in August 2025, Perplexity shipped a software update within 24 hours to circumvent it. Judge Chesney singled that move out specifically in her ruling. Hard to argue good faith when you're speed-running around a block the day it goes up.

Amazon backed up its claims with more than just legal theory — it spent over $5,000 in engineering hours building new detection systems to filter Comet's automated ad traffic. Security researchers at Brave had already flagged prompt-injection vulnerabilities in Comet back in October 2025, and enterprise testing found the browser more susceptible to phishing than standard Chrome. Amazon folded those findings into its original complaint.

One wrinkle worth sitting with: Amazon's founder Jeff Bezos is a personal investor in Perplexity. And AWS signed a $38 billion infrastructure deal with OpenAI on November 3, 2025 — the day before Amazon sued Perplexity. Whether that timing is coincidence or chess, you decide.

Is This Really About Security — or About $68.6 Billion?

Here's what the security framing conveniently buries: Amazon pulled in $68.6 billion in advertising revenue in 2025. Every one of those dollars depends on a user traveling through a search results page stuffed with sponsored listings before reaching checkout. An AI agent that skips directly to purchase doesn't just bypass the buy button — it erases the entire ad layer that sits between a search and a sale. Perplexity made this exact argument in a November blog post, calling Amazon's campaign "bullying" and noting that agentic shopping would theoretically generate more transactions for Amazon, not fewer. That framing landed less well in court than it did on the internet.

Amazon CEO Andy Jassy said in November that agentic commerce "has a chance to be really good for e-commerce" but argued that AI agents aren't yet accurate enough on personalization and pricing. Convenient position for a company building its own agentic shopping tools — which Amazon is. It has separately blocked ChatGPT from shopping on its platform too. Meanwhile, the Amazon Alexa.com launch this week placed Alexa+ directly alongside Gemini, Claude, and ChatGPT as a browser-based AI assistant, extending it beyond smart speakers for the first time.

What Does This Mean for AI Agents and the Open Web?

Amazon updated its Business Solutions Agreement on March 4, 2026, adding a formal requirement that all AI agents identify themselves when accessing its services. If this injunction survives appeal, it sets a precedent that platforms can deny access to AI agents even when the user explicitly authorized the action. That's a significant clause to add to the internet's operating rules.

The deeper legal question — whether the Computer Fraud and Abuse Act applies to agentic software executing tasks on a human's behalf — has never gone to trial. This case puts it squarely on the docket. The Ninth Circuit gets to decide whether a user's permission is enough, or whether every platform in between gets its own veto.

Perplexity says this is about user choice. Amazon says it's about security and platform integrity. Both of them have revenue motives they'd rather not discuss too loudly.