In one of the most telling legal confrontations of the AI era, Amazon has filed a lawsuit against the San Francisco-based startup Perplexity AI, alleging that its new “Comet” tool violates Amazon’s terms by autonomously browsing and making purchases on the platform.
At first glance, this may look like a typical intellectual-property or access-control dispute. But beneath the surface lies a deeper tension between two visions of the digital economy: one dominated by tightly controlled platforms, and another shaped by intelligent, autonomous agents acting directly on behalf of users.
The Case That’s Stirring the AI Market
Filed on 5 November 2025 in the U.S. District Court for the Northern District of California, the lawsuit (Amazon.com Services LLC v. Perplexity AI Inc.) accuses Perplexity of enabling “unauthorised automated interactions” with Amazon’s marketplace.
Amazon’s complaint describes Comet as an AI-powered shopping assistant that can log into users’ accounts, search for products, compare prices and even place orders — all while disguising itself as a regular human user. The company claims this behaviour breaches its Terms of Service, which prohibit the use of “robots, spiders or other data-gathering tools” to access its site.
It also argues that Comet creates consumer-protection and cybersecurity risks, interferes with Amazon’s quality-control systems and threatens the integrity of its marketplace. According to Business Standard’s report, Amazon issued multiple cease-and-desist notices before resorting to legal action, alleging that Perplexity deliberately bypassed blocks meant to stop Comet from functioning.
Perplexity, on the other hand, calls the lawsuit “corporate intimidation.” In statements to The Guardian and The Hindu, the company defended Comet as a privacy-respecting agent that never stores user credentials on its own servers. It insists that Comet only acts on behalf of users who have granted explicit permission , and that Amazon’s attempt to block it amounts to an attack on user autonomy.
Platform Power vs. User Agency
This isn’t just a question of who can automate online shopping. It’s about whether the internet will evolve into a closed ecosystem policed by major platforms, or an open environment where users , through their chosen AI agents , can act freely.
Amazon’s position represents the platform-centric model that has defined the digital economy for two decades: control access, mediate every interaction, and monetise every user action. Perplexity’s Comet embodies a newer paradigm , one where AI agents become independent intermediaries between humans and the web.
If Amazon wins, it reinforces the notion that users can only interact with online services through the interfaces and APIs authorised by the platform. If Perplexity prevails, it could open the door for a generation of “agentic” AI tools capable of performing real-world tasks from ordering groceries to booking travel without direct user clicks.
The clash therefore isn’t about scraping data or violating a user agreement. It’s about who gets to control the flow of digital labour: the platform, or the person represented by their AI agent.
The Legal Grey Zone of ‘Agentic’ AI
Legally, the dispute sits in uncharted territory. Traditional computer-access laws were built around human actors using automated scripts or bots. But when an AI agent acts as a true extension of a human, using their credentials, following their intent, and making real purchases, it blurs the boundary between “unauthorised access” and “delegated authority.”
Can a company claim that a user’s own AI assistant is trespassing on their account? Does a platform have the right to deny access to a digital proxy acting for a legitimate account holder?
These are not hypothetical questions. As generative and agentic AI systems evolve, they will increasingly perform tasks once reserved for humans , managing finances, executing contracts, and negotiating prices. Yet, as this case shows, the law hasn’t yet decided whether such agents are independent entities, or simply “extensions of the user’s will.”
What’s at Stake?
For Amazon, the case is existentially tied to its business model. A shopping agent that can skip promoted listings and paid placements directly threatens the economics of Amazon’s marketplace. Its algorithms, advertising architecture, and recommendation engines all depend on keeping the user within its controlled environment.
Allowing third-party agents to autonomously navigate and purchase products could dismantle that model, reducing visibility for sponsored products and breaking the data feedback loops that drive Amazon’s advertising revenue.
For Perplexity, this is about survival and principle. The startup has positioned Comet as the next step in human-AI collaboration: an assistant that doesn’t just suggest, but acts. In an open letter, the company accused Amazon of “trying to outlaw the future of personal AI,” arguing that users have the right to choose how they interact with online platforms , whether through a mouse, voice command, or AI agent.
The Broader Policy Vacuum
The case also exposes the vacuum in global AI governance. Regulators from Brussels to Washington are drafting laws on AI transparency, bias, and accountability but few frameworks address autonomous economic agents.
Who regulates an AI that spends money, negotiates deals, or enters into transactions? Should such systems be required to identify themselves to online platforms? And when they act on behalf of a user, who bears the responsibility if something goes wrong, the agent’s developer, the user, or the platform hosting the transaction?
These questions will soon dominate digital-policy discussions. If the court recognises agentic AI as a legitimate user proxy, it may push regulators to create new categories of “digital agents” with defined rights and obligations. If not, platforms could be empowered to gatekeep AI access, limiting innovation to those who play by their proprietary rules.
What Comes Next?
Legal analysts expect the case to turn on two points:
- Whether Comet’s automated behaviour qualifies as a breach of Amazon’s Terms of Service or a violation of anti-circumvention laws; and
- Whether a user’s authorisation can legitimise an AI’s access, even if the platform disapproves.
A narrow ruling might focus only on contractual compliance. A broader one could set precedent on digital personhood, user autonomy, and the very definition of “access” in the age of machine intermediaries.
Either way, Amazon v. Perplexity will be remembered as one of the first courtroom tests of agentic AI. It’s not just about shopping carts, it’s about who truly controls the digital economy when machines start to act on our behalf.
Read the cease and desist letter by amazon to perplexity here.