Amazon has accused Perplexity’s AI browser, Comet, of slipping past defenses to access private customer accounts, and sent a cease and desist to stop it. Perplexity fired back, calling the move an attack on user freedom and automation. The fight could define what agentic browsing is allowed to do, with implications for OpenAI’s Atlas and any AI that navigates the web for you.
- Amazon says Comet posed as a human to bypass controls and reach private accounts.
- Perplexity argues Amazon is trying to lock down automated shopping and limit user choice.
- The outcome could shape the rules for agentic AI across major platforms.
- Two out of three consumers still do not trust AI to make purchases on their behalf.
Why did Amazon move against Perplexity now?
Amazon drew a sharp line. In a cease and desist letter, the company alleges Perplexity’s Comet did more than browse public pages. It reportedly disguised itself as a human user to evade safeguards and access private customer accounts without permission. If true, that veers into unauthorized access land and raises fresh alarms about data security and privacy.
Amazon put it bluntly in the filing. The dispute, they say, is about an automated agent that “disguised” itself to get into accounts. An Amazon spokesperson added that third-party agent apps must follow the same rules as everyone else and said the company repeatedly asked Perplexity to remove Amazon from Comet because it “significantly” degrades the shopping and support experience.
Recommended Tech
Concerned about unauthorized access? Consider a digital security service like Aura to help protect personal data and finances across your accounts.
How is Perplexity defending its approach?
Perplexity isn’t backing down. In a public response, the company framed Amazon’s position as a monopolistic attempt to control automated shopping. The argument is philosophical as much as legal. For decades, software was a passive tool. With agentic AI, software acts like labor. It can be an assistant, an employee, an agent.
Kent Byrnes, Perplexity’s vice president of legal, was direct. Amazon, he said, wants a monopoly over automated shopping agents. To Perplexity, the fight is about user freedom and the future of internet commerce.
What is agentic browsing and why does it matter?
Agentic browsers such as Perplexity’s Comet and OpenAI’s Atlas use large language models to carry out multi-step tasks on your behalf. They can search, click, fill out forms, compare prices, and, in some cases, check out. Think of them as tireless personal shoppers that can read, decide, and act online.
That power creates friction with platform rules. As analyst Sara Mitra of Luna Base AI notes, “These agentic AIs challenge the traditional boundary between user automation and violation of platform policies, especially regarding personalization and data security.” For a deeper dive, see Luna Base AI’s technical analysis.
Get the latest tech updates and insights directly in your inbox.
Could OpenAI’s Atlas and other agents be next?
Observers wonder whether Amazon’s action is a one-off or the start of broader enforcement across marketplaces with strict access controls. Stanford tech-law specialist David Lin says if Amazon’s argument that undisclosed bots are unauthorized gains traction, other agentic agents, including those from OpenAI, may face similar scrutiny.
Perplexity already faces a separate suit from Reddit over alleged stealthy agent behavior. The legal questions reach into classic areas like the Computer Fraud and Abuse Act, as explored by this overview. However it lands, platforms and AI makers will be forced to clarify what bot behavior is allowed and what crosses the line.
Are consumers ready to let AI buy for them?
Not quite. A February 2025 Omnisend survey found that 66 percent of consumers will not allow AI to make purchases for them, even with the promise of better deals. The trust gap remains wide. E-commerce analyst Glenna Moore puts it plainly. Mainstream adoption hinges on trust and transparency, and current lawsuits probably slow that timeline.
What happens next in the agentic AI fight?
This is bigger than one cease and desist. As legal analysis around the case points out, the outcome could set the rules of the road for how AI agents operate on commercial sites. Luna Base AI argues Amazon’s push to enforce control through legal means may reset expectations for all online automation.
In the near term, expect platforms to tighten bot detection, AI companies to add clearer disclosures and controls, and regulators to watch closely. Longer term, the market will reward those who can balance user choice, platform integrity, and privacy.
Recommended Tech
As agentic AI workloads grow, local AI performance matters. Consider AI-ready PCs like the Lenovo IdeaPad Slim 3X AI Laptop from the Copilot+ generation to run more tasks on-device.
Frequently asked questions
What exactly did Amazon allege in its letter?
Amazon claims Perplexity’s Comet masqueraded as a human user to bypass protections and access private customer accounts without Amazon’s permission. Amazon also says it asked Perplexity multiple times to remove Amazon from the Comet experience.
Did Perplexity admit to accessing private accounts?
No. Perplexity disputes Amazon’s framing and argues that the company is trying to monopolize automated shopping. Perplexity positions agentic AI as a legitimate tool that users should be free to deploy.
Is agentic browsing legal?
It depends on the conduct and the platform’s rules. Using automation transparently and within published terms is generally safer. Undisclosed access to private areas, evasion of technical barriers, or misrepresentation can raise legal risks under computer misuse laws and contracts.
Could OpenAI’s Atlas face similar pushback?
Possibly. Legal experts say if the theory that undisclosed bots are unauthorized access holds, other agentic AIs that operate on gated commerce platforms could see similar challenges.
Can website owners block agentic bots?
Yes. Site owners can use authentication, rate limits, bot detection, CAPTCHAs, and clear terms prohibiting automated access. They can also issue legal notices when they believe access violates those terms.
Should I let an AI agent make purchases for me?
Most people are not ready. Surveys show a strong trust gap. If you try it, start small, set strict spending limits, require confirmations before checkout, and review transaction history closely.
