Amazon has sent Perplexity a cease and desist letter, requesting it stop allowing people to use its AI agents to buy things.
getty
Amazon has fired warning shots across the bow of AI agents. Specifically, the giant e-commerce company has served notice to Perplexity, the AI-powered answer engine, requesting that it stop allowing people to create and use purchasing agents from its Comet browser to buy things on Amazon’s website.
Stop for a moment and read that back: an e-commerce vendor doesn’t want people buying things from its website. At least, not the way those people want to. Rather, the e-commerce vendor wants people to buy things from it in the way it wants them to.
And there you have the heart of the issue: it’s about power.
Amazon’s throwing the legal equivalent of a fairly significant fit, as is clear from how it describes the use of web browsing technology in the cease-and-desist letter that is now public: “Perplexity must immediately cease using, enabling, or deploying Comet’s artificial intelligence (‘AI’) agents or any other means to covertly intrude into Amazon’s e-commerce websites,” Amazon lawyer Moez M. Kaba writes, adding that such “intrusions” are, in Amazon’s view, violations of both federal and state computer fraud and abuse laws.
Welcome to the battle for the future. Of commerce, to be sure, but much more than that as well.
Let’s unpack what’s happening here:
- You create an AI agent with Perplexity
- You ask it to buy something on Amazon for you
- It does so, but doesn’t identify itself as an agent to Amazon
From Perplexity’s perspective, this is just the next evolution in technology and automation. And Amazon’s dislike of agentic AI purchasing isn’t about the actual purchases as it is about the loss of control over how you find products.
“Amazon should love this,” Perplexity states a blog post titled “Bullying is Not Innovation.” “Easier shopping means more transactions and happier customers. But Amazon doesn’t care. They’re more interested in serving you ads, sponsored results, and influencing your purchasing decisions with upsells and confusing offers.”
Perplexity says that agentic AI is a transition phase for technology from being a simple wrench or hammer to being an actual assistant or even employee. Amazon says it’s fine with innovation, AI and agents, writing in the cease-and-desist letter that “Amazon shares the industry’s excitement about AI innovations and sees significant potential for agentic AI to improve customer experiences in a range of areas.”
But there’s a snag. That snag is transparency.
Perplexity’s agent – or is it your agent if you’re using it to buy something – does not identify itself as an AI agent. Rather, it logs in with your credentials from your own personal device, in effect logging in as you.
The offline equivalent is a parent sending a kid to the store for a product. The kid buys it, comes home and all is well. But from Amazon’s perspective, the kid needs to identify as an agent of the parent.
Perplexity begs to differ. “User agents are exactly that: agents of the user,” the company writes. “They’re distinct from crawlers, scrapers, or bots. A user agent is your AI assistant—it has exactly the same permissions you have, works only at your specific request, and acts solely on your behalf.”
In other words, it’s essentially you. And it doesn’t need to identify as an agent, according to Perplexity.
The concern from Perplexity is likely that if its agents identify as such, Amazon could block them. Or it could change product pricing dynamically. Or it could add a service fee. The concern from Amazon is that agentic commerce takes user experience, shopping experience, context and impulse purchases out of the picture.
“Amazon has invested billions of dollars over many years to develop a carefully curated shopping experience in the Amazon Store,” the cease and desist says. “That shopping experience is designed to help customers find and discover products that cater to
their needs based on key elements, including reviews, price, availability, delivery speed, measures of post-purchase satisfaction like return rates, and each customer’s own browsing and shopping history. This delights our customers and earns their trust, which is critical to the success of the Amazon Store.”
That’s of course a vastly different perspective than author and activist Cory Doctorow shares on Amazon in his recent book Enshittification, which suggests that Amazon abuses both users and merchants and that “top results in a product search aren’t the best matches: they’re the matches that pay the highest fees to be top of the list.”
Be that as it may, all of this seems likely to be tested in court.
“This is the first major legal test of autonomous AI agents in commerce,” says Lumida Wealth Management, an investment advisory company.
It goes beyond that, of course. As we increasingly employ agents to do work for us, we’ll also have to explore who owns that work, whether we can submit that work to employers as our own, and whether those agents, in acting on our behalf, have the same rights and privileges and responsibilities that we do.
“This isn’t just about shopping bots,” says Hashbyt, a U.K. software company. ”It’s about the foundation of an AI-driven web.”