Amazon sues Perplexity for shopping AI agency; OpenAI's security chief: technology is still in early stages and poses risks.

Amazon, the American e-commerce giant (Amazon), has officially filed a lawsuit against the AI startup Perplexity, accusing its Comet AI browser and the automated shopping agent feature “agentic shopping” of accessing Amazon user accounts without permission and disguising automated actions as human behavior, resulting in security risks and a compromised user experience.

E-commerce giant Amazon has sued the AI browser Perplexity.

According to reports, Amazon pointed out that Perplexity secretly logged into Amazon customer accounts through its Comet browser and AI agent CometAI, disguising automated activities as real human browsing. Amazon stated that this action raised security concerns and that they had repeatedly requested to stop but were ignored. Amazon emphasized that the behavior of CometAI not only undermined the personalized shopping experience it had built over many years but also interfered with the advertising and recommendation services Amazon provides to consumers.

Perplexity previously stated that Amazon had issued legal threats to prevent Comet AI from shopping on its platform. Perplexity called this action a threat to user choice and the future of AI assistants. They stated on their official blog: “When large companies use legal threats and intimidation to stifle innovation and make people's lives worse, that is bullying.”

Amazon itself has also developed shopping agents like Buy For Me and Rufus.

Perplexity is one of many startups attempting to redefine the browsing experience with AI, aiming to create an automated agent capable of handling everyday online activities such as writing emails and completing shopping. Its agentic technology can place orders, compare prices, and even manage daily online activities for users. The rise of such AI assistants has sparked controversy over the legitimacy of website authorization and automated behavior.

Amazon itself is also developing similar tools, including the Buy For Me feature that allows cross-brand shopping within its app, as well as the AI assistant Rufus for recommending products and managing the shopping cart.

Shopping agents are still in the early stages, OpenAI warns of risks.

In contrast to Amazon's defensive attitude, Walmart announced last month a collaboration with OpenAI that allows ChatGPT users to complete purchases directly within the conversation without having to enter the official website, calling this move agentic commerce in action (.

The so-called agentic AI browser is still an early technology, and there are known potential issues. OpenAI also acknowledges that its ChatGPT Atlas may mistakenly represent users in purchasing wrong products. OpenAI's Chief Security Officer Dane Stuckey stated, “The ChatGPT agent functionality is powerful and designed to be safe, but it can still make mistakes, such as purchasing the wrong products or forgetting to confirm with the user before executing critical actions.”

This article discusses Amazon suing Perplexity for shopping AI agents, and OpenAI's Chief Security Officer stating that the technology is still in its early stages and poses risks. It first appeared in Chain News ABMedia.