The e-commerce giant accuses the startup of fraudulent access and violation of its terms of use
Amazon sent a letter of formal notice to Perplexity AI via the law firm Hueston Henningan, demanding that the startup immediately stop using its chatbot Comet to make purchases on its platform. The Seattle group accuses the company founded by Aravind Srinivas of having “disguised” its agent as a Chrome browser to bypass its detection systems, in violation of Computer Fraud and Abuse Act (CFAA) and its Californian equivalent.
According to Amazon, Perplexity would have “refused to operate transparently” despite several warnings since November 2024, and would have “taken active steps to conceal” its activities in the Amazon Store. The document, dated October 31, 2025 and signed by lawyer Moez M. Kaba, cites “unauthorized intrusions,” “harm to customer experience” and “data security risks.”
A legal battle with economic implications
Amazon bases its complaint on the CFAA, a law regularly used against third-party actors who have continued to access computer systems after a blockage or formal notice, as in the case Facebook vs. Power Ventures. The giant cites a “significant cost” for its security teams, forced to track down and block Comet’s access, before Perplexity bypasses them again.
The company also accuses the AI agent of “altering the customer journey” by ignoring optimization criteria (price, availability, speed of delivery, personalized suggestions) and by “risking distorting consumer choices”. Amazon believes that these practices “undermine customer trust and the integrity of the Amazon model.”
Perplexity’s response: the right to choose your AI
In a message published on X, Perplexity denounces an intimidation tactic.
“We believe that every user has the right to use the artificial intelligence of their choice. Amazon wants to prevent Comet users from using an AI assistant to purchase on its platform, and it is now using legal threats to do so. But we will not be intimidated. »
CEO Aravind Srinivas defends the legitimacy of using an agent mandated by a user, believing that “it is not up to Amazon to monitor the way in which a customer delegates an action”.
On X, he adds:
“We would be happy to work with Amazon to find a win-win outcome. But if Amazon tries to block our Comet assistant and harm our users, we will have to defend ourselves and not be intimidated. »
A conflict at the heart of the new agentic economy
The case crystallizes the question of the interoperability of AI agents. Amazon wants to impose that any agent operates in a transparent and identified manner on its platform, while Perplexity claims the principle of user delegation, with the idea that an agent acts legally in the name of the person who has mandated him.
This debate opens a strategic fault line.
- For Amazon, controlling the transaction means protecting customer trust and the value of its advertising model based on search placement.
- For Perplexity, allowing an AI agent to act freely means guaranteeing freedom of use and innovation.
This conflict could become the first major legal precedent of the era of AI agents. If it went to court, it could define the limits of the right of access and digital delegation on major platforms.
Beyond the dispute between Amazon and Perplexity, it is the question of control of the automated act of purchasing that looms, to what extent can artificial intelligence act on behalf of a user? A central issue for the future web, which should set a precedent and relaunch discussions on the responsibility of AI agents.
