Researchers warn about Open AI's browser

TT

Published 12.02

 

Sam Altman, CEO of Open AI. Archive photo. Photo: Jeff Chiu/AP/TT

The new AI browser Atlas is criticized for lack of security.

Experts that the tech site The Verge spoke to warn that fraudsters can exploit simple tricks.

American Open AI, the company behind Chat GPT, launched a new browser just under two weeks ago. The service, Atlas, is built around the company's cash cow Chat GPT.

Unlike a more traditional browser, users can let Chat GPT navigate and search through sites and tabs. The service can also, for example, find, book and pay for flights or make medical purchases.

But the fact that Atlas has so much AI integrated into the service is met with criticism.

“They know more about you and are more powerful than regular browsers,” Yash Vekaria, a researcher on online privacy and security at the University of California, tells The Verge.

Vekaria warns that attackers can sneak instructions into the AI ​​part of Atlas and thus make it do things that the user does not want, such as changing the delivery address for goods or sharing saved sensitive information. Some methods of instructing AI are as simple as writing white text on a white background on a site.

Several of the experts The Verge spoke to warn that AI is still unable to distinguish between a scam site and a real one, for example when it comes to financial fraud.

A professor at the University of Kent in the UK urges people to only use AI “when they really need it” and that browsers should be “AI-free” by default.