Ninth Circuit Clears Perplexity's Comet in Amazon CFAA Case

Introduction

On August 4, 2026, the U.S. Court of Appeals for the Ninth Circuit vacated a preliminary injunction that had barred Perplexity AI's Comet browser from operating its AI shopping assistant on Amazon.com. Writing for a three-judge panel in case No. 26-1444, Judge Milan D. Smith Jr. held that Perplexity does not itself "access" Amazon's servers under the Computer Fraud and Abuse Act (CFAA) — the human users who direct Comet do. It is the first federal appeals court decision to squarely address how legal liability for an AI agent's actions should be split between the company that built the agent and the person who told it what to do, and it lands at a moment when browser- and shopping-agent products from multiple companies are racing to automate exactly the kind of behavior Amazon tried to block.

Feature Overview

The case began after Amazon obtained a district court injunction blocking Comet from browsing and purchasing on Amazon.com on users' behalf, arguing the practice violated the CFAA and California's parallel computer-crime statute. On appeal, the Ninth Circuit panel held that Amazon was unlikely to succeed on the merits of that claim, because the CFAA's "access" requirement attaches to whoever directs the action, not to the software that carries it out. Even where Perplexity's systems receive a user's Amazon account information to carry out instructions, the court held that this operational role falls short of the control needed to treat the developer, rather than the user, as the party who "accessed" Amazon's servers. The panel explicitly acknowledged there is "little to no existing caselaw directly dealing with how to ascribe responsibility for AI agents" under the statute, and applied the rule of lenity — the doctrine that ambiguity in a criminal statute must be read in the defendant's favor — to resolve that gap in Perplexity's favor. Critically, the court described its holding as narrow: it addresses only the CFAA and state-law computer-fraud "access" question. Amazon's trademark infringement claims and other state-law theories survive and continue in the U.S. District Court for the Northern District of California, where the case now returns for further proceedings.

Usability Analysis

For Comet users, the practical effect is immediate: the browser can resume directing purchases and account actions on Amazon.com while the underlying case proceeds, reversing the restriction that had been in place since the district court's original injunction. For Perplexity, the ruling removes an existential threat to one of Comet's flagship capabilities without resolving the litigation outright — the trademark and remaining state-law claims still have to be litigated on the merits, and Amazon has signaled it may seek en banc rehearing or petition the Supreme Court. For the wider agentic-browsing category, including OpenAI's ChatGPT Atlas, Google's AI Mode and Project Mariner, and Microsoft Copilot's agent mode, the decision offers a legal framework, though not a guarantee, that agents requiring explicit per-action user direction are more likely to be treated as tools carrying out a user's own access rather than as independent intruders. The panel was careful to tie its reasoning to the degree of user direction involved, which means agents built to act with less explicit, step-by-step instruction from a human may not benefit from the same analysis.

Pros and Cons

Pros:

• Establishes an early appellate precedent distinguishing agent developers ("tools") from the users directing them ("principals") under the CFAA, reducing legal exposure for agent-based products
• Prevents a preliminary injunction from shutting down an AI agent's core functionality before the underlying legal questions are fully litigated on the merits
• Gives the fast-growing category of browser and shopping agents — not just Comet — clearer, though still evolving, legal footing to keep operating

Cons:

• The holding is explicitly narrow and could still be reversed on en banc rehearing or a future Supreme Court appeal, leaving long-term legal risk unresolved
• Amazon's trademark infringement and remaining state-law computer-fraud claims survive and continue in district court, so the underlying dispute is far from over
• The ruling ties its reasoning to explicit, per-action user direction and offers no clear standard for more autonomous agents that act with less direct step-by-step instruction, which will likely invite further litigation

Outlook

Amazon's next moves, whether it seeks en banc rehearing at the Ninth Circuit or petitions the Supreme Court, will determine how durable this precedent turns out to be, and the remanded trademark and state-law claims mean the Amazon-Perplexity dispute itself is not finished. More broadly, the ruling arrives as AI companies across the industry push agents deeper into e-commerce, email, and other third-party services that were never designed with autonomous software in mind. Courts, platforms, and lawmakers are now working out, largely case by case, where the line falls between a legitimate user-directed tool and unauthorized automated access. The panel's own words point to more change ahead: "the legal understanding of agentic AI will doubtless change." Platform operators like Amazon may respond by leaning more heavily on technical countermeasures or updated terms of service rather than CFAA litigation, since this ruling narrows, without closing, that legal avenue.

Conclusion

This ruling gives Perplexity's Comet, and by extension the broader field of user-directed browsing and shopping agents, real breathing room to keep operating while the underlying legal fight over platform access continues. It is a meaningful, closely watched data point, not a final resolution, since the trademark claims proceed separately and Amazon retains further appeal options. For developers and companies building agentic products that touch third-party platforms, the case is worth tracking closely as an early signal of how U.S. courts are likely to allocate responsibility between AI tools and the people who use them. Rating: 4/5, for the clarity it brings to an unsettled area of law while leaving real questions about more autonomous agents unanswered.

Editor's Verdict

Ninth Circuit Clears Perplexity's Comet in Amazon CFAA Case earns a solid recommendation within the it news space.

The strongest case for paying attention is establishes an early appellate precedent distinguishing agent developers from the users directing them under the CFAA, reducing legal exposure for agent-based products, which raises the bar for what readers should now expect from peers in this space. Reinforcing that, prevents a preliminary injunction from shutting down an AI agent's core functionality before the underlying legal questions are fully litigated adds practical value rather than just headline appeal. The broader signal worth registering is straightforward: the Ninth Circuit's August 4, 2026 ruling in case No. 26-1444 is the first federal appellate decision to directly address CFAA liability for an AI agent acting on a user's behalf. On the other side of the ledger, the holding is explicitly narrow and could still be reversed on en banc rehearing or a Supreme Court appeal is a real constraint, not a marketing footnote, and it should factor into any serious decision. Layered on top of that, amazon's trademark infringement and remaining state-law computer-fraud claims survive, so the underlying dispute is far from resolved narrows the set of teams for whom this is an obvious yes.

For AI industry watchers, strategy teams, and decision-makers tracking platform shifts, this is a serious evaluation candidate, not just a curiosity to bookmark. For everyone else, the safer posture is to monitor coverage and revisit once the use cases that matter to your team are demonstrated in the wild.