A U.S. appeals court has handed Perplexity a major victory, lifting a judicial order that had temporarily prevented the AI company from deploying its automated shopping tools on Amazon’s marketplace.
In a closely watched dispute that could influence the future of agent-driven artificial intelligence, the Ninth Circuit Court of Appeals concluded that Amazon is unlikely to prevail on its claim that Perplexity’s technology violates federal anti-hacking laws.
The case has become one of the first major legal tests over whether AI agents—software capable of browsing websites, making purchases and carrying out tasks for users—can legally interact with online platforms on a person’s behalf.
Amazon expressed disappointment with the ruling, saying it disagrees with the court’s decision to overturn the preliminary injunction and is considering its legal options. Perplexity, meanwhile, described the outcome as a win for consumer choice, maintaining that internet users should be free to decide which AI systems they want to use.
The legal battle began last year when Amazon accused the startup of secretly accessing customer accounts through its Comet browser and related AI assistant. According to Amazon, the technology could log into shopping accounts and complete transactions for users, creating potential security concerns. The company also alleged that Perplexity ignored multiple requests to halt the activity.
Perplexity rejected those allegations, arguing that the lawsuit was designed to restrict customers from using competing AI tools. The company contended that its software merely acts on behalf of users and that Amazon was attempting to protect its own commercial ecosystem.
A California federal judge sided with Amazon earlier this year, issuing a temporary order that blocked Perplexity’s shopping agents from operating on the platform. The judge had found that Amazon presented substantial evidence suggesting the AI tools may have violated federal computer fraud statutes.
The appeals court reached a different conclusion.
Its ruling emphasized that the individuals using Perplexity’s software—not the company itself—were the ones accessing Amazon’s systems. Because the users had authorized the AI agents to act on their behalf, the court said Amazon’s argument under federal computer-access laws was unlikely to succeed at this stage.
Beyond the immediate dispute between the two companies, the decision carries broader implications for the rapidly expanding world of agentic AI. Developers across the technology industry are racing to build systems capable of planning tasks, comparing products and executing transactions with minimal human intervention.
As those tools become more common, courts may increasingly be asked to determine where user authority ends and AI responsibility begins. Tuesday’s ruling marks an early chapter in that debate, but it is unlikely to be the last.
