Amazon vs. Perplexity: When a Cease-and-Desist Letter Calls Your AI a “Computer Fraud”
Amazon sending a cease-and-desist letter to Perplexity over its Comet AI shopping agent reads like a niche spat between Big Tech and a hot startup.
Underneath it, though, is a very clean, very teachable case study in:
- how to weaponize Terms of Service, computer-fraud theories, and “user experience” into a single high-leverage demand letter;
- how quickly a demand letter can escalate into full-blown litigation if the other side refuses to blink; and
- what any SaaS/platform/AI company should be doing now to avoid receiving an Amazon-style letter of its own.
What Comet Is, and Why Amazon Cares
Perplexity’s Comet is an AI-powered browser that lets users delegate web tasks to an “agentic” assistant: it reads, summarizes, and, crucially for this story, logs into sites and makes purchases on the user’s behalf. (Wikipedia)
For e-commerce, that means a user can say: “Buy me X on Amazon,” and Comet will, using the user’s own credentials stored locally on their device, navigate Amazon, compare options, and complete checkout. Perplexity’s marketing pitch is simple: more convenience, less clicking. (Perplexity AI)
Amazon sees something very different:
- automated traffic that does not identify itself as a bot,
- a shopping flow that bypasses Amazon’s ad placements and recommendation systems, and
- a third party inserting itself between Amazon and “its” customer relationship.
Amazon publicly stated that tools like Comet can “significantly degrade the shopping and customer service experience,” and that third-party agents must operate transparently and respect Amazon’s rules. (About Amazon)
That framing is important, because it’s exactly the theme of the cease-and-desist letter.
The Cease-and-Desist: What Amazon Actually Alleged
According to coverage of the letter and Amazon’s own statements, the cease-and-desist accuses Perplexity of:
- violating Amazon’s Terms of Service by using automated agents (essentially “robots”) to browse and purchase without disclosure, and
- engaging in “computer fraud” by disguising Comet’s automated activity as if it were human shopping. (Futurism)
Amazon’s standard conditions prohibit “data mining, robots, or similar data gathering and extraction tools” unless expressly authorized. Comet’s entire value proposition is to act as an autonomous robot that looks like a human in Amazon’s logs. (SiliconANGLE)
Amazon also reportedly told Perplexity that:
- Comet’s behavior degrades the customer experience on Amazon,
- introduces privacy and security risks by funneling traffic through an AI agent, and
- had already been the subject of repeated private requests to stop, which Perplexity ignored or circumvented. (SiliconANGLE)
From a demand-letter perspective, this is exactly what you want on the sender’s side: a clear contractual hook, articulated harms that sound bigger than “technical scraping,” and a record of prior warnings.
How Perplexity Framed the Letter: “Bullying, Not Innovation”
Perplexity answered with a very public blog post titled “Bullying is not innovation.” (Perplexity AI)
In Perplexity’s telling, the situation looks like this:
| 🧑💻 Perplexity’s Story | 📦 Amazon’s Platform |
|---|---|
| Users are simply using “their own” AI assistant (Comet) as a personal shopper, just like a human assistant. | Amazon is being forced to interact with an automated agent it never agreed to and that violates express ToS prohibitions on bots. |
| Credentials stay on the user’s device, not on Perplexity’s servers, so there is no new data risk. | The browsing and purchase flow still comes from an undisclosed automation layer, raising security, integrity, and UX concerns. |
| Amazon wants to protect its ad-driven business model and control over upsells, not customers. | Amazon says this is about transparency, security, and its right to decide which automated agents can transact on its platform. |
Perplexity characterizes the letter as an “aggressive legal threat” aimed at preventing users from choosing their own AI assistants, analogizing software to a wrench or an employee: a tool the user has the right to employ. (Perplexity AI)
That public response is effectively a counter-narrative to Amazon’s demand letter: instead of quietly complying, Perplexity is trying to re-litigate the framing in the court of public opinion.
From Cease-and-Desist to Lawsuit: The Escalation
Amazon didn’t stop at the demand letter. Within days, it filed suit in federal court (Northern District of California), alleging that Perplexity’s Comet:
- covertly accessed customer accounts,
- disguised automated browsing as human activity, and
- posed security risks by bypassing Amazon’s safeguards and personalization systems. (Reuters)
Press reports on the complaint quote language that sounds very much like the letter’s rhetoric, including lines to the effect of:
“Perplexity’s misconduct must end… that Perplexity’s trespass involves code rather than a lockpick makes it no less unlawful.” (Reuters)
From a demand-letter practice standpoint, this is instructive. The cease-and-desist isn’t just an angry email; it’s the pre-complaint narrative that Amazon then copies into its pleading. The allegations, theories, and even the turns of phrase carry over almost verbatim.
Legal Theories Hiding Inside the Letter
The parties haven’t published the full text of the cease-and-desist, but between Amazon’s statement, coverage of the letter, and the subsequent complaint, you can reverse-engineer the main theories. (About Amazon)
| ⚖️ Theory / Cause of Action | 💡 How Amazon Uses It in the Narrative |
|---|---|
| Breach of Contract (ToS) 📜 | Perplexity’s agents violate explicit prohibitions on robots, data mining, and undisclosed automation. |
| Computer Fraud / CFAA 💻 | Automated agent masquerades as human browsing, allegedly “trespassing” into areas Amazon has told it not to access. |
| Trespass to Chattels / Interference 🧱 | Automated traffic interferes with Amazon’s systems and the curated shopping experience Amazon built. |
| Unfair Competition / Interference with Business Relationships 🧩 | Comet disrupts Amazon’s relationship with its customers and its advertising/personalization revenue streams. |
The letter is not trying to win the entire CFAA debate in one shot. It’s doing something narrower and more practical: putting Perplexity on notice that Amazon considers this conduct unlawful access rather than “just another browser,” and that if Perplexity continues, Amazon will tell exactly that story to a judge.
For any platform or SaaS founder, this is a reminder that your ToS is not just boilerplate; it is the first exhibit in your demand letter and complaint when you claim someone is misusing your site.
Demand-Letter Craft in Amazon’s Approach
Look at Amazon’s moves through the lens of demand-letter strategy, and a pattern emerges.
Building the Record Before the Letter
Amazon reportedly:
- warned Perplexity repeatedly about Comet’s behavior, including a prior promise by Perplexity to stop,
- blocked certain automated access routes, and
- made clear (internally and externally) that bots must identify themselves and respect platform rules. (SiliconANGLE)
By the time the cease-and-desist goes out, Amazon can credibly say: “We tried to solve this quietly; you agreed and then circumvented us.” That is powerful context in both a letter and a lawsuit.
Framing Conduct as Both Technical and Human Harm
The allegations aren’t limited to “bot bad.” Amazon blends:
- technical infractions (undisclosed automation, violation of robots/data-mining bans);
- user harm (degraded recommendations, inaccurate pricing/delivery context, risk of wrong items); and
- systemic harm (threatening the integrity of its checkout and personalization systems, which hold immense commercial value). (SiliconANGLE)
That mix makes it easier for Amazon to argue for strong remedies and for a court to see this as more than a petty platform turf war.
Positioning the Letter as the Last Exit Before Litigation
Given how quickly the lawsuit followed, the cease-and-desist looks less like “let’s see if they’ll talk” and more like “final warning, we’ll file next.”
This is a common pattern in serious commercial disputes: you draft the letter with a view toward (1) being able to attach it to the complaint if you need to, and (2) creating a record that the defendant knowingly proceeded after notice.
What This Fight Means for Other Platforms and SaaS Companies
Comet is an extreme example, but the logic applies to far more mundane products:
- browser extensions that auto-add coupons or re-rank products;
- price-comparison widgets that ride on top of other stores’ websites;
- scraping tools, reseller dashboards, automation scripts that manage multiple accounts;
- AI agents that fill carts, manage bookings, or navigate dashboards.
Amazon’s message is simple: if you automate interactions with our site in a way we’ve prohibited, hide that automation, and degrade the experience we design, we will treat you as a trespasser and a fraudster, not as “just another client.” (SiliconANGLE)
For platforms, that should translate into some concrete action items:
| 🏗️ Area | ✅ Healthy Posture After Amazon v. Perplexity |
|---|---|
| Terms of Use | Explicit bot/agent clauses, clear disclosure requirements, and reserved rights to block non-compliant automation. |
| Technical Controls | Rate limiting, bot detection, and a plan for handling automated agents that ignore ToS. |
| Notice and Takedown | A defined contact path for agents and integrators who want to be “good citizens”; escalation procedures for those who don’t. |
| Communications Playbook | Template demand letters ready to go when someone crosses the line, plus an internal decision tree for when to escalate to suit. |
If you don’t have those, the Amazon letter is basically a free checklist.
Lessons for Businesses That Rely on AI Agents
Even on the user side, this fight has implications.
Perplexity’s defense leans hard on the idea that Comet is just “your own assistant” acting on your behalf, with credentials stored locally and no extra data risk. (Perplexity AI)
But researchers have already demonstrated serious security issues with agentic browsers, including Comet, such as prompt-injection attacks that trick the agent into leaking credentials or making scam purchases on fake sites. (Windows Central)
If you are a business letting AI agents log into your operational accounts—Amazon, banks, SaaS dashboards—this should worry you on two levels:
- Platform risk: the site owner might decide your agent of choice violates its ToS and suddenly block the workflow you rely on.
- Security risk: the agent can be manipulated into doing things a human user would spot as malicious immediately.
That is precisely the kind of risk profile that leads to demand letters and litigation when something goes wrong.
Takeaways for Demand-Letter and Contract Practice
From a lawyer’s standpoint, Amazon vs. Perplexity is almost a case study you could hand to associates on “how to draft a high-impact commercial demand in an AI context.”
A few core takeaways:
- ToS is your first weapon. When you write a company’s online terms, assume you might one day rely on them to call someone’s automated behavior “unauthorized access” or “computer fraud.” Draft accordingly.
- Narrative matters as much as doctrine. Amazon does not rely solely on the fact that its ToS say “no robots.” It paints a story of degraded user experience, privacy risk, and systemic harm—then attaches legal labels (CFAA, trespass, breach) to that story.
- Demand letters should be drafted with the complaint in mind. The language in the subsequent lawsuit strongly echoes the themes in the cease-and-desist, showing that Amazon’s lawyers were already thinking about the pleading when they wrote the letter. (Reuters)
- Public counter-narratives are part of the game now. Perplexity’s decision to respond with a blog titled “Bullying is not innovation” shows that sophisticated recipients will sometimes litigate the optics immediately. Your letter needs to be written with that possibility in mind.
For clients on either side—platforms sending letters, startups receiving them—the Amazon–Perplexity clash is a preview of the new normal in AI-driven commerce: demand letters that treat bots as trespassers, ToS as the primary shield and sword, and public messaging as part of the legal strategy rather than an afterthought.