Perplexity AI Fights Amazon in Court That Will Decide If a 1986 Law Can Stop AI Agents

The judges didn't hide how strange the case felt. U.S. District Judge John Hinderaker, sitting on the Ninth Circuit panel by designation, said the 1986 Computer Fraud and Abuse Act was "not really built" to consider AI. "Does an AI agent ever have intent?" he asked. It's a question Ronald Reagan's Congress never anticipated — but the answer could determine whether AI agents remain legal to deploy on the open web.

On Thursday, the U.S. Court of Appeals for the Ninth Circuit heard oral arguments in Amazon.com Services LLC v. Perplexity AI Inc., according to Reuters. The case is one of the first to consider who is liable for the actions of agentic AI — a fast‑growing sector where systems plan, reason, and execute tasks with limited human oversight.

The three‑judge panel wrestled with how to map AI behavior onto a statute written when 2400‑baud modems were cutting edge. Bloomberg Law reported that the panel pushed back on Perplexity's argument that its Comet AI agents gain authorized access to Amazon through users — similar to how Safari or Chrome act as a browser for people.

U.S. Circuit Judge Eric Tung framed the core question as one of analogies. "The user is giving the key to Perplexity, and Perplexity is then entering Amazon's servers or computers," he said, according to Reuters. "Much of this case turns on the proper analogy." Is Comet a browser with extra steps, or an intruder that happens to have the homeowner's key?

Amazon sued Perplexity in November 2025, accusing the startup of covertly accessing private Amazon customer accounts through its Comet browser and associated AI agent. Comet can log into users' online shopping accounts and place orders on their behalf. Amazon said Perplexity's system posed security risks and ignored repeated requests to stop.

Perplexity fired back that the lawsuit was a "bald attempt" to block Amazon users from using Comet because AI agents "don't have eyeballs to see the pervasive advertising Amazon bombards its users with."

In March 2026, a federal judge temporarily blocked Perplexity from using its agent on Amazon, finding "strong evidence" that the startup violated the CFAA, according to Jones Day's analysis. The judge relied on the 2016 Facebook v. Power Ventures precedent, which held that a platform can revoke third‑party access even when users share their credentials. The Ninth Circuit stayed the injunction pending appeal.

This isn't just about one shopping bot. The Ninth Circuit's ruling will shape the legal boundaries for every AI agent that interacts with third‑party websites — from research assistants that scrape public data to coding agents that pull from package registries to autonomous assistants that book travel or manage finances.

The central question: is an AI agent's access "unauthorized" under the CFAA when a platform revokes permission, even if individual users have consented? Under Power Ventures, a cease‑and‑desist letter can trigger CFAA liability. But that case involved a social media aggregator, not an AI system acting on a user's behalf.

Perplexity told the appeals court in its May 2026 appellate brief that Amazon's CFAA theory was "a fundamental misfit" for an AI agent that users explicitly direct. Legal experts told RabbitRank the decision could "define the legal framework" for agentic AI for years.

If the Ninth Circuit sides with Amazon, any website could block AI agents by sending a cease‑and‑desist letter — and the developer deploying that agent could face federal hacking charges. That would fundamentally change how agents interact with the web, forcing builders to negotiate access agreements with every site their tools touch.

If Perplexity wins, the precedent could establish that AI agents are extensions of user intent — not independent actors — and that users have the right to deploy AI tools to interact with services they're already authorized to access. Either outcome will shape the infrastructure decisions of every company building agentic AI products.

As the IAPP noted, the dispute highlights unanswered questions about AI agent liability — when an AI agent does something wrong, is the user responsible? The developer? The model provider? The courts haven't decided yet.

Judge Hinderaker posed that question during Thursday's oral arguments,.¹ The Ninth Circuit has not indicated when it will rule. The case is No. 26‑1444, with Amazon represented by Hagan Scotten of Hueston Hennigan and Perplexity by Chris Michel of Quinn Emanuel Urquhart & Sullivan. A decision could come within weeks or months.

Meanwhile, AI agents aren't waiting for the courts to catch up. Companies across the industry are racing to deploy agentic systems, from coding assistants that manage multi‑file projects to research agents that browse the web autonomously. The legal risks exist whether or not the Ninth Circuit clarifies them — but the ruling will at least tell builders which side of the 1986 law they're standing on.