Amazon demands Perplexity shut down its AI shopping browser, sparking a clash over who controls your online purchases.
Amazon has escalated its dispute with Perplexity by sending a cease and desist letter over the startup's Comet browser, which uses AI agents to make purchases on behalf of users without requiring them to visit Amazon's website or manually enter payment details. The e-commerce giant accuses Perplexity of violating its terms of service and committing computer fraud by failing to disclose when the AI is actively shopping. According to Amazon, it had already asked Perplexity to stop using AI agents on its platform a year ago, and the startup agreed. That agreement fell apart when Comet launched.
In tests, Comet proved remarkably efficient at completing transactions. When asked to buy a product, the browser accessed the user's existing account information and placed an order in about 30 seconds, using the default payment method and address without requiring additional authentication.
This seamless experience is precisely what troubles Amazon. According to PC Mag, the company argues that third-party applications offering to make purchases on behalf of customers should operate transparently and respect whether businesses want to participate. Amazon also worries that bypassing its website devalues the platform and eliminates opportunities for browsing, upsells, and ad exposure.
Perplexity fired back with a pointed statement, accusing Amazon of bullying and using litigation to stifle innovation. The startup argues that easier shopping should benefit Amazon by driving more transactions and creating happier customers. Instead, Perplexity contends that Amazon prioritizes serving ads and sponsored results over genuine customer convenience.
The company framed the dispute as a matter of principle, suggesting that Amazon doesn't believe users have the right to employ AI assistants to act on their behalf. Observers have noted the tension underlying this clash: Amazon wants users to interact with its own AI shopping tools, not competitors' versions.
In terms of community insights, opinions vary. Agentic AI browsers remain nascent technology with documented vulnerabilities. OpenAI has acknowledged that its ChatGPT Atlas can make mistakes, including purchasing the wrong product or forgetting to confirm actions.
Security experts warn of prompt injection attacks, where malicious instructions embedded in websites or emails could trick an AI agent into ignoring user commands and buying unintended items instead. Perplexity's response largely sidesteps these safety concerns, instead emphasizing the transformative potential of AI labor and assistants. Note that these topics also show up in modern cybersecurity certifications. The widening AI landscape has created new challenges and threats within the cybersec community.
And the dispute highlights diverging corporate strategies. While Amazon is taking a hardline stance, (according to a corporate release from the company) Walmart recently partnered with OpenAI to allow ChatGPT users to shop its catalog through Instant Checkout technology, demonstrating that agentic commerce can work when both parties agree. As this technology matures, the question is who gets to decide how they operate and whose interests they serve.
