ChatGPT Atlas is the new browser on the block, but it comes with myriad security concerns

OpenAI recently launched a new web browser called ChatGPT Atlas, and it’s unlike any browser most of us have used before. Instead of just letting you surf the web, Atlas brings ChatGPT, the AI chatbot, right into the browsing experience. This means you can ask questions about web pages, get summaries, or even automate tasks like managing subscriptions or shopping online. It’s a mix of a browser and a personal assistant, all in one.

Sounds exciting but security experts and privacy advocates are already raising concerns, according to a Washington Post report. One of the main reasons is the way Atlas remembers what you do online. The browser has a “memory” feature that keeps track of the websites you visit, what you click on, and even actions you take on those sites. OpenAI says it doesn’t store extremely sensitive info like passwords or medical records, but tests have shown that it can remember things like a user’s registration for health services. In other words, some private information could be stored without users realizing it.

Managing this memory isn’t very easy either. You can tell Atlas not to remember certain sites, but it’s not very intuitive. Simply clearing your browsing history doesn’t always remove the stored memories, leaving users with little control over what the browser keeps track of.

Another big concern comes from the AI agents that Atlas can use to perform tasks automatically. These agents can log into your accounts, make purchases, and manage subscriptions. While this is convenient, it also creates risk. Giving an AI agent access to login credentials and payment information makes it a potential target for mistakes, glitches, or even hacking. Even with safeguards, the risk can’t be completely eliminated.