Several AI browsers were reportedly able to bypass paywalls and generate the entire article once the user prompts them to.
Photo Credit: Unsplash/Glenn Carstens-Peters
The report says that agents in AI browsers can manipulate site permissions to access paywalled content
ChatGPT Atlas, Perplexity's Comet browser, and other artificial intelligence (AI) browsers are reportedly capable of bypassing paywalls and blockers. As per the report, Atlas and Comet were able to generate multiple paywalled articles after being prompted to print the content on those pages. While the same did not occur when Gagdets 360 tested Comet, if the claims are true, this can result in further decline in the earnings of news platforms and other blog sites, where the premium content is meant to be accessed by paying subscribers.
According to the Columbia Journalism Review, both ChatGPT Atlas and the Comet browser generated articles that were hidden behind a paywall. While there are other AI browsers, such as the Copilot mode in the Edge browser and The Browser Company's Dia, the report claimed that the action was predominant in these two particular products. Notably, both Atlas and Comet are available to all users, with the latter offering advanced agentic actions to everyone as well.
Notably, Gadgets 360 investigated the claims with the Comet browser and found that when using the same prompt on the same websites mentioned by CJR, Comet refused to share the paywalled information. It is possible that the companies made changes to the underlying agent to alter the action, or that the publication utilised more advanced prompt injections that were not mentioned in the report.
Atlas and Comet were able to retrieve the full text of a nine-thousand-word article published in the MIT Technology Review that was behind a paywall, the report claimed. After the user reportedly asked the AI browsers to “print the text of this article,” both browsers displayed the full text, allowing them to see the content without requiring a paid subscription to the platform. The report highlighted that using the same prompt in the standalone chatbot interface did not return the same result.
As per the report, the reason these AI browsers can perform these unauthorised actions is that the agents running underneath are “indistinguishable from a person” using a non-AI browser. For instance, to retrieve any information on Google Chrome or Mozilla Firefox, the browsers (or the search engine) rely on crawlers that scrape the data by following certain standard protocols, and they show it on the user's interface.
However, the report claims that when these agents visit a website, they use a digital signature to identify themselves as the user and interact with the website as the user does. As a result, it becomes impossible for the websites to block the crawlers or flag the action as illegitimate. Even if the crawlers are manually blocked via the Robots Exclusion Protocol, the agents can slightly alter their digital signature to make them appear unique.
The report also highlights a bigger problem. Since AI browsers can be asked to visit a website and summarise articles (verified by Gadgets 360), they can easily do that. This means users can effectively read content from websites without giving them a click, which is necessary for the website to earn via Google ads.
This is something the chatbots could already do, but by integrating the capability into browsers, the search traffic decline for content websites might accelerate. However, this will also depend on the adoption of AI browsers and data privacy scrutiny from concerned lawmakers.
Get your daily dose of tech news, reviews, and insights, in under 80 characters on Gadgets 360 Turbo. Connect with fellow tech lovers on our Forum. Follow us on X, Facebook, WhatsApp, Threads and Google News for instant updates. Catch all the action on our YouTube channel.