ChatGPT debuted in November 2022. The chatbot built by OpenAI has been a favourite with everyone from the very beginning. Now the company has introduced its web browser Atlas. This browser is said to be a rival to Google Chrome! But this time, experts have issued a warning for users. A recent study has found that this browser is not as completely secure. Experts from ‘LayerX Security’ have claimed that hackers can easily take over the browser using secret codes or commands and users will not even notice!

The report says, ‘Atlas does not yet use any strong anti-phishing protection system. Which means that users are 90 percent more likely to be attacked by hackers than traditional browser users like Chrome or Edge.’

And in this case, the feature called ‘Memory’ plays an important role. It is known that this feature was introduced by OpenAI in February 2024. This feature remembers minute details of users, such as names, preferences, interests, etc. And this is what hackers are taking advantage of. Once they get access to this feature, they just have to turn off the user’s device. Or log in later. Hackers get their work done with the help of that leaked information.

ChatGPT has written a new success story in just three years with over 800 million people use it. In this situation, this new web browser Atlas is being seen as a new source of income. Especially, efforts will be made to increase income through advertising and online search. But this time, a cloud of fear has gathered around that browser.

Why Open AI Atlas can be a threat to users?

The main threat posed by the OpenAI Atlas browser is its vulnerability to indirect prompt injection attacks, which can be exploited by malicious websites to hijack the browser’s integrated AI agent. This design collapses the traditional separation between data and instructions, allowing hostile commands hidden in web content to be executed with the user’s authority.

Unlike traditional browsers, which rely on security mechanisms like the same-origin policy, Atlas’s AI agent has broad privileges and can interact with all logged-in accounts and browser tools. This gives attackers a new, more dangerous attack surface.

An attacker can use a hidden prompt to command the browser’s AI agent to steal sensitive information from other accounts the user is logged into, such as email, banking, or cloud storage. The browser’s AI agent can be tricked into performing unauthorized actions on the user’s behalf. Examples include transferring funds, sending emails, or accessing personal files.

A compromised AI agent could inject malicious links into summaries of web pages or spread disinformation. To function effectively, the AI browser needs to store a comprehensive memory of the user’s browsing habits, search queries, and login sessions. This data becomes a valuable target for attackers.

Atlas AI browser exposes the user to a “potential attack vector”

AI browsers fundamentally change the browser’s security model by treating untrusted web content as executable instructions. This exposes the user to a “potential attack vector” that did not exist in traditional browsers. Because the AI agent operates with the user’s full authority, it can bypass standard browser protections like sandboxing and same-origin policies that prevent malicious code from accessing other accounts. Even with safeguards implemented by OpenAI, security researchers have demonstrated that it is possible to bypass them. The core security vulnerability remains a “frontier, unsolved security problem,” and new attacks are likely to emerge as fast as companies can patch them.

Long or Short, get news the way you like. No ads. No redirections. Download Newspin and Stay Alert, The CSR Journal Mobile app, for fast, crisp, clean updates!

App Store – https://apps.apple.com/in/app/newspin/id6746449540

Google Play Store – https://play.google.com/store/apps/details?id=com.inventifweb.newspin&pcampaignid=web_share