Major digital news publishers in the US have backed Amazon in its lawsuit against Perplexity, saying unauthorised AI agent access to content should be blocked.

Amazon sued Perplexity in November, arguing the AI start-up was accessing its shopping website and user accounts without permission.

A California judge temporarily blocked Perplexity from accessing Amazon’s website via a preliminary injunction issued in March.

Perplexity is now appealing that decision. It has argued that its AI agents are “much more transparent and limited” than Amazon’s own use of agentic AI to complete transactions on third-party retailer sites.

Amazon alleges that Perplexity has “purposely configured” Comet AI so that when the AI tool is deployed on behalf of an Amazon customer, “Perplexity falsely identifies its Comet AI agent activity as coming from Google Chrome, which is a separate, widely used web browser owned by Google. As a result, Perplexity’s Comet AI agent covertly poses as a human customer shopping in the Amazon Store on a Google Chrome browser.”

It said this poses “considerable risks” to Amazon customers including making them vulnerable to cyber attacks.

Amazon said it told Perplexity executives at least five times between November 2024 and October 2025 that their AI agents may not covertly access its store but it continued to happen.

Digital Content Next (DCN) has filed an amicus brief in support of Amazon, meaning it is not a direct party to the lawsuit but has strong interest in the case.

It said it objects to Perplexity’s exploitation of protected content for the AI company’s own commercial benefit.

DCN members include many major US media organisations including the Associated Press, BBC Studios, Bloomberg, Conde Nast, Dow Jones, the Financial Times, Fox News, The Guardian, Hearst, The Independent, NBC Universal, News Corp, The New York Times, Paramount, NPR, Vox Media and The Washington Post.

DCN told the court that “giving AI agents unregulated access to protected content and systems threatens public access to information by undermining investments in journalism”.

DCN told the court: “DCN’s news publisher members invest immense resources to create high-quality journalism and take steps to protect that content.

“If publishers cannot effectively prevent access by AI agents deployed by third-party commercial actors, publishers will be unable to protect the income streams that support journalism. In the current digital landscape, content providers need to be able to secure access to their content in order to attract advertisers and grow the number of paid digital subscriptions.”

AI agents and human subscribers being blurred means ‘harder time selling ad space’

The group said that “unauthorised agentic AI access to user account information and other protected content threatens advertising revenue in several major ways. To state the obvious, advertisers pay for their ads to be seen by humans, not AI agents.

“If publishers cannot distinguish between an AI agent and a human subscriber, they will not be able to measure audiences accurately.”

It said this would cause them to have a “harder time selling ad space”, citing the example of digital publisher Salon which lost a mid-sized ad buyer because they were concerned about the scale of non-human visitors.

“As this example illustrates, unrestricted AI agent access thus forces publishers into a difficult choice: either sink valuable time, money, and focus into technical countermeasures to detect and filter out non-human ad impressions, or allow undisclosed AI agents (and thus those AI agents’ parent companies) silently to access (and appropriate for their own uses) publisher content without viewing or engaging with the advertisements that fund the creation of such content.

“Either way, publishers will be unfairly forced to pay for AI agents’ attempts at free access.”

DCN went on: “If other AI firms’ agents are permitted to interface with publishers’ protected web assets the same way as Perplexity’s Comet AI agent interfaces with Amazon, it will be even harder for publishers to measure actual human audience size, and thus their entitlement to ad revenue, accurately – at least not without being drawn into an open-ended ‘arms race’ to stay ahead of agentic AI developments that make those agents more and more difficult to detect.”

DCN added that publishers “rely on data from user activity to develop sophisticated marketing strategies that attract new subscribers” but AI agents intruding could make this data “unreliable or unusable”.

It also said that if an AI agent gets through the door using one subscriber’s login, it could then “extract all content to which that subscriber has access, and then summarise, repackage, strip of attribution, and distribute it” for the profit of its parent company.

DCN went on to argue that all publishers (not just newsbrands) are entitled to control access to their content and that being forced to open the “gates” to AI agents, even those that have obscured their identities and purpose, could affect their security and their ability to negotiate a fair price to access or license their content.

It said the content Perplexity procures from publishers through its Comet AI agent might otherwise cost it millions of dollars in content licensing deals. Perplexity has to date done several deals with the likes of Getty, Gannett, Le Monde, The Independent, Los Angeles Times and Time.

“An AI agent that commercialises publisher content for AI products without authorization effectively commandeers a perpetual license of unlimited scope without paying a cent to the publisher who created the content, and thereby deprives the publisher of an important, regular source of revenue. This is simply not sustainable.”

Chris Pedigo, DCN’s senior vice president of government affairs, told members in a memo: “At a high level, the case examines whether AI systems and user agents can bypass publisher safeguards to access content under false pretences – without transparency or in violation of terms – and extract and repurpose that content in ways that may undermine its value.”

He added: “This case sits squarely at the centre of one of the most consequential questions facing our members: how trusted content is accessed, valued, and monetised in an AI-driven ecosystem.

“At issue is not simply unauthorised scraping, but a broader shift toward agentic AI systems that bypass publisher controls, obscure their identity, and extract high-value content and data at scale… these practices risk undermining the core economics that support investment in premium content – distorting advertising signals, weakening subscription models, and eroding the direct relationships between publishers and their audiences that are foundational to long-term growth.”

Email pged@pressgazette.co.uk to point out mistakes, provide story tips or send in a letter for publication on our “Letters Page” blog