Amazon’s legal threat to Perplexity raises questions over AI autonomy and platform control

AI startup Perplexity has accused Amazon of using legal threats to block innovation after the e-commerce giant demanded that its Comet browser stop allowing AI agents to shop on Amazon on behalf of users.

In a blog post titled “Bullying is Not Innovation,” Perplexity described the move as an attack on user choice and a warning sign for the future of agentic AI.

The dispute highlights growing tensions between dominant online platforms and emerging AI tools that aim to act autonomously for users, raising new questions about competition, compliance, and control in digital commerce.

Perplexity said Amazon’s demand threatens the idea of “user agents,” which it defines as AI assistants that act with the same permissions as the user and perform tasks only on their behalf.

It also accused Amazon of putting ad revenue ahead of user convenience, citing CEO Andy Jassy’s recent comments to investors about advertising returns.

Amazon has disputed that characterization, saying its actions are intended to protect customers and ensure service quality. In a statement, the company said that third-party applications that make purchases on behalf of users must operate transparently and respect businesses’ choices about participation.

“Agentic third-party applications such as Perplexity’s Comet have the same obligations, and we’ve repeatedly requested that Perplexity remove Amazon from the Comet experience, particularly in light of the significantly degraded shopping and customer service experience it provides,” Amazon added.

Platform control vs. AI autonomy

Analysts say the dispute highlights the practical and commercial challenges that arise as AI agents begin to operate directly on behalf of users.

“The legal threat shows that the future of agentic AI is not as seamless as the industry perceived,” said Lian Jye Su, chief analyst at Omdia. “Despite open standards like MCP and A2A, most brands and applications value direct, unique user traffic to their platforms and interfaces as the main revenue drivers. As such, they do not appreciate seamless integration with other AI agents that now sit between unique users and these applications, thereby reducing unique user traffic and potentially traffic-based revenue from advertising and usage.”

Leslie Joseph, principal analyst at Forrester, said Amazon’s action can be read as “an opening salvo in a broader fight for control of the interface.” Agentic browsers like Perplexity’s Comet act as brokers between users and storefronts, stripping away the ads, recommendations, and pricing tactics that underpin Amazon’s margin model.

“This shifts the locus of influence from platform to consumer,” Joseph added. “The episode is a pointer to what’s coming: dominant platforms are on the back foot defending their ‘closed’ ecosystems, while agentic browsers are trying to reopen the web around user-directed automation.”

Amazon’s own initiatives in AI-powered shopping add another layer to the conflict. The company is developing services such as “Buy For Me” and the “Rufus” assistant, which can recommend and purchase products within its ecosystem. Amazon’s move to restrict third-party AI agents may also serve to safeguard these internal projects and their associated revenue streams.

Navigating agentic AI risks

The outcome of the Amazon–Perplexity dispute could help shape future rules governing how AI agents and online platforms interact. Analysts say it may eventually lead to clearer frameworks covering access controls, user authentication, data exchange, and revenue-sharing between agents and applications.

“However, this will have further challenges because not all applications monetize in the same way, and the revenue models vary significantly across verticals,” Su added. “So there may not be a one-size-fits-all solution even with legal resolutions.” 

For enterprises, the case serves as a reminder to deploy agentic AI tools with greater caution.

“It’s not a stretch to imagine that agents that rely on scraping, browser automation, or gray-zone access will face barriers as platforms tighten control over APIs and data channels,” Joseph said. This means that enterprises must take extra care when their AI agents interact with competitors’ or vendors’ IP, user data, and revenue systems. “In many cases, such instances happen unknowingly, so this requires enterprises to fully understand their AI agent behaviors and put in the right framework to minimize conflicts with competitors and other vendors,” Su added.