The clash between Amazon and Perplexity marks a turning point for AI-powered commerce. By obtaining the legal blocking of Comet, the American giant is not only targeting an automated purchasing agent, but it is defending its control over access to its platform and its users' data. Behind this decision rendered in San Francisco, a broader question emerges: to what extent can AI agents act on behalf of Internet users without encroaching on the sovereignty of large platforms?

In brief
- Amazon has obtained a preliminary injunction against Perplexity, for now blocking its AI browser Comet from making purchases on its platform on behalf of users.
- American justice has not yet decided the merits of the case, but at this stage it validates Amazon's argument that a user's agreement is not enough to authorize an AI agent to access its services.
- The conflict is part of an older confrontation between the two groups, with Amazon claiming to have increased its warnings before taking legal action.
- This decision could set a precedent, by more clearly defining the extent to which an AI agent can act on behalf of a user on a large platform.
Amazon gets first stop against Comet
Amazon has won its first legal victory against Perplexity in a case that directly affects the growth of agentic commerce, while the firm has just strengthened its ties in the field of AI. Indeed, federal judge Maxine Chesney granted a preliminary injunction preventing, at this point, the Comet browser from making purchases on Amazon on behalf of users.
The decision does not yet resolve the substance of the dispute, but it already sets out a central point. For the court, the authorization given by a user to an AI agent does not automatically constitute authorization given by the platform itself.
This sequence is part of an older confrontation between the two groups. Amazon claims to have issued multiple warnings before taking legal action and claims that Perplexity continued its operations despite technical blocking measures. The immediate issue is therefore not only commercial. It also concerns the ability of a platform to control automated access to its services, especially when they pass through protected customer accounts.
- Federal Judge Maxine Chesney granted a preliminary injunction against Perplexity, an interim measure and not a final judgment;
- According to the decision, Amazon presented qualified elements “generally uncontested evidence” to argue that Perplexity accessed password-protected Prime accounts with users' consent, but without Amazon's permission;
- Amazon says it has asked Perplexity to shut down at least five times since November 2024;
- After a technical blockage put in place in August 2025, Perplexity would have deployed an update within 24 hours to circumvent the obstacle;
- The order requires Perplexity to cease this access and destroy copies of Amazon customer data collected through Comet. His application was suspended for seven days to allow an appeal.
A case that exposes the fragilities of agentic commerce
The second reading of the file is broader. It concerns the place that platforms are prepared to leave to agents capable of purchasing, navigating and arbitrating in the customer's place. Perplexity has placed its defense in the field of user rights.
The company says it wants “continue to defend the right of Internet users to choose the AI they want to use”. In a blog post published in November thanks to on-chain data, she had already described Amazon's offensive as intimidation, while maintaining that agentic commerce could generate more transactions for the Seattle giant.
This line resonates with a previous statement from Andy Jassy, CEO of Amazon, according to which agentic commerce “could be very beneficial for e-commerce”while remaining, according to him, insufficiently reliable in terms of personalization and prices. Amazon does not contest the idea of commerce driven by AI per se, but the fact that a third party is installed between its marketplace and its users without its green light.
The other side of the matter concerns the safety and economics of the model. Amazon relied on work by Brave published on October 21, 2025, which found that Comet had prompt injection vulnerabilities via screenshots and malicious web content.
Brave writes that these attacks show that traditional web security assumptions “no longer hold when AI agents act on behalf of users” and that this type of helper can be manipulated by untrusted content running with the user's authenticated privileges.
At the same time, Amazon is also defending a massive economic issue. The group generated $68.6 billion in advertising revenue in 2025, and an agent who jumps directly from request to payment erases the sponsored placements that structure part of the platform's monetization. The industrial context adds an additional layer. AWS and OpenAI announced a multi-year, $38 billion strategic partnership on November 3, 2025, with targeted capacity deployment before the end of 2026.
Beyond the setback inflicted on Perplexity, the rise of AI agents will come up against the red lines set by the platforms. It remains to be seen whether this framework will be further tightened with new regulatory or political initiatives, such as the decree “Genesis Mission”now scrutinized as a possible signal of acceleration.
Maximize your Tremplin.io experience with our 'Read to Earn' program! For every article you read, earn points and access exclusive rewards. Sign up now and start earning benefits.
