The new ChatGPT Atlas browser marks a radical shift: browsing becomes a two-way conversation with AI. The assistant can summarise pages, compare prices, and even complete purchases on behalf of users. “You’re literally giving the mouse to AI,” says digital strategist Jaap van Zessen, who tested the tool. “It’s incredibly useful—but also risky if you don’t realise what data it’s accessing.”

For AI expert Jarno Duursma, Atlas represents more than convenience. “It’s part of a larger power struggle over the internet’s infrastructure,” he says. “Tech giants want full control—from energy supply to social media to browsers. OpenAI isn’t just chasing Google—it wants to own the gateway itself.”

Privacy experts are alarmed. Cybersecurity analyst Sijmen Ruwhof found that Atlas may transmit entire webpages, including logged-in content, to OpenAI servers. “That’s far beyond what Chrome does,” he notes. “Imagine your company portal or online banking pages being summarised by a remote AI.” Although OpenAI claims to strip personal identifiers like IDs or payment details, Ruwhof warns that centralising all browsing data under one company remains a major vulnerability.

Security concerns extend beyond data collection. Atlas’s autonomous browsing feature could be exploited by attackers inserting hidden commands into websites, tricking the AI into sharing sensitive information. Duursma remains cautious: “Chrome has two decades of testing behind it—Atlas has just launched. ‘Move fast and break things’ might not be the best principle for something handling your private life.”

Still, the allure is strong. By automating online routines, Atlas could redefine how users interact with the web. As Van Zessen puts it, “Those who grow up with this will see the old internet as obsolete.” But between convenience and control, the line is getting thinner.