A federal appeals court is about to decide whether the AI agent you direct to shop for you is acting as your tool or as an intruder. On June 11, 2026, a three-judge panel of the Ninth Circuit heard oral arguments in Amazon v. Perplexity, a case over whether Perplexity's agentic browser Comet violates the Computer Fraud and Abuse Act when it logs into a user's Amazon account, browses, and completes purchases on that user's behalf. No ruling came from the bench, and a decision is likely months away. But the question at the heart of the case is one of the most consequential in tech right now: when a human delegates a task to an AI agent, who counts as the authorized visitor, the person or the software?

  • The Ninth Circuit heard oral arguments in Amazon v. Perplexity on June 11, 2026; no ruling has been issued.
  • Amazon sued Perplexity in November 2025, alleging its Comet agent accessed Amazon accounts without authorization under the CFAA and California law.
  • A district judge granted Amazon a preliminary injunction in March 2026; the Ninth Circuit then stayed it pending appeal.
  • The ruling will set the first federal precedent on whether buyer-side AI agents can access third-party platforms at a user's direction.

What actually happened

Amazon sued Perplexity AI in November 2025, alleging that Comet, Perplexity's AI-powered browser, accessed Amazon's systems without authorization in violation of the federal Computer Fraud and Abuse Act and California's computer-fraud statute. Comet can log into a user's Amazon account using the user's own stored credentials, browse products, and check out on the user's behalf. In March 2026, District Judge Maxine Chesney granted Amazon a preliminary injunction, finding Amazon likely to prevail and concluding that Comet's access was not authorized by Amazon, regardless of any permission the user granted. About a week later, the Ninth Circuit paused that injunction pending appeal, letting Comet keep operating while the case proceeds. At the June 11 oral argument in Seattle, Circuit Judges Milan Smith Jr. and Eric Tung, joined by a district judge sitting by designation, pressed both sides on how to map AI behavior onto a statute written when 2400-baud modems were cutting-edge.

RelatedChrome Puts an AI Agent at the OS Level on 200 Million Phones

Why does a 1986 law decide the fate of AI agents?

Because the Computer Fraud and Abuse Act, passed in 1986, is the primary US law governing unauthorized computer access, and it was written long before anyone imagined software that acts autonomously on a person's behalf. The core dispute is about the word "authorized." Perplexity argues that when a user directs Comet to use their own credentials and access their own account data, the agent is simply the user's tool, no different in principle from Safari or Chrome acting as a browser for a person. Amazon argues something closer to platform sovereignty: a private company is entitled to set the rules for access to its systems, and an automated tool that bypasses those rules is unauthorized even if the user consented, because the platform's consent is also required. Judge Tung captured the tension with an analogy at argument, framing it as the user handing a key to Perplexity, which then enters Amazon's computers. Whether that handoff is authorized access or unauthorized intrusion is the entire case, and the 1986 statute never contemplated it.

The mechanism most coverage skips

The reason this case matters far beyond shopping is that the same legal question governs nearly every AI agent that touches a third-party system. A research assistant that scrapes public data, a coding agent that pulls from a package registry, a travel agent that books flights, a finance assistant that logs into your bank: all of them rely on accessing platforms that did not explicitly authorize an automated visitor. If the Ninth Circuit sides with Amazon, the platform-consent theory could expose the developers and even the everyday users of these tools to civil or criminal liability under the CFAA, because the platform, not just the user, would need to permit the agent. The ACLU filed an amicus brief warning that such a ruling could chill not only commercial agents but journalism and public-interest research that depend on automated access. This is why a shopping-bot dispute drew that kind of attention. The court is effectively deciding the default rule for the entire agentic web: are AI agents presumptively allowed to act for the humans who direct them, or do they need each platform's blessing first?

Who this affects

AI agent developers face the most direct exposure, because a ruling for Amazon would force them to navigate a web where every platform can legally block automated access regardless of user consent, fragmenting what agents can actually do. Platforms like Amazon gain a powerful tool to control or exclude agents from competitors if the platform-consent theory prevails, which has obvious competitive implications. Users sit in the middle: a win for Perplexity affirms that the tools they direct can act for them, while a win for Amazon could make many agent features legally risky to offer. And the broader internet gets its first real precedent on a question that will define the next era of how software interacts with the web, one that affects everyone from startups to researchers to ordinary people who increasingly hand tasks to AI.

RelatedFirefox Bets on the Opposite of Everyone Else: an AI Off Switch

What is next

Watch for the Ninth Circuit's opinion, which could take months and will be the precedent everyone in the agent space reads closely. Watch whether the court rules narrowly on the specific facts of Comet and Amazon or broadly on the authorization question, because the breadth of the holding determines how much it reshapes the agentic web. Watch how platforms adjust their terms of service and technical defenses in anticipation, since many are already weighing how to handle agent traffic. And watch Congress and regulators, because a CFAA written in 1986 straining to govern 2026 technology is exactly the kind of mismatch that eventually prompts legislative attention, whatever the court decides.

Our take

This is one of the most important tech cases in years, and it is flying under the radar because it looks like a fight about a shopping bot. It is not. The Ninth Circuit is being asked to set the default rule for whether AI agents are extensions of the people who direct them or trespassers that need each platform's permission, and that choice will shape what the agentic web is allowed to become. The user-as-principal view that Perplexity advances is the one that preserves user agency and lets the technology develop; the platform-sovereignty view that Amazon advances hands incumbents a gatekeeping power that could smother competing agents in the name of access control. We think the user who hands over their own key and their own credentials is exercising a right, not committing a crime. But the law here is genuinely unsettled, the statute is decades out of date, and a ruling either way will echo across every AI agent built to act on a person's behalf.

Reporting via Search Engine Journal, analysis by GenZTech.