AI browsers are here. Should you care?
By Jared Newman | Published: 2025-10-31 18:58:00 | Source: Fast Company – technology
Earlier this week, I had AI handle all my grocery shopping.
using Comet Perplexity BrowserI provided a link to my shopping list on Google Keep, then asked him to put everything in my shopping cart for the Kroger pickup order, making sure to select previous purchase items when multiple options were available. Within a few minutes, Comet selected all the right items — including the taco shells and fake meat we typically get on taco night — and put me on the check-out page.
This kind of scenario explains why many AI companies are now trying to create their own browsers. Guilty of confusion Dia Browser CompanyBoth became widely available without invitation earlier this month, and OpenAI launched its own program ChatGPT Atlas Browser for MacOS last week. Opera started previewing the An AI-powered browser called Neonand Microsoft Edge includes Pilot “Copilot Actions” feature It can browse the web for you.
All of these browsers promise to automate the mundane aspects of navigating the Internet, and there are moments — like tackling your grocery list — when that seems pretty convincing. But these AI browsers also make some serious compromises in security, privacy, and ease of use that make me wary about using them.
This story first appeared on advisorJared’s weekly newsletter about tech tips subscription Get more tips every Tuesday.
What is an AI browser, anyway?
Most people don’t think about web browsers much, but they are very important programs. Replacing Chrome with a more privacy-oriented browser like Brave or DuckDuckGo, for example, can dramatically reduce the amount of data companies collect about you online, while powerful user browsers like Vivaldi can help you deal with tab overload. (See my Browser domain directory¤For more details.)
Meanwhile, AI browsers aim to eliminate the friction of using AI. With ChatGPT Atlas, for example, the default search box leads to ChatGPT instead of Google, and there’s an “Ask ChatGPT” button that answers questions about the current page in the chat sidebar. You can also request details from other open tabs or search pages from your browsing history.
Unless you’re constantly using AI, some of these things may seem unnecessary. Opening ChatGPT or Claude in a browser tab (or using dedicated mobile and desktop apps) is not Which Cumbersome, and you can summarize a web page just by including a link in your query.
But what makes AI browsers really interesting is their ability to interact with your personal context.
For example, I manage all my bookmarks in Raindrop.io. While reviewing my bookmarks for technology news in ChatGPT Atlas, I asked for one-line summaries with links to each item, similar to the weekly reports in My newsletter. I still have to rewrite each summary, but Atlas at least got me started and saved me time copying and pasting links into the text. This was only possible because Atlas was logged into Raindrop and could see the contents of my bookmarks list.
More ambitiously, these browsers promise to navigate through web pages automatically. AI companies call this “proxy” browsing (a term that makes me sick for reasons I can’t fully express).
Getting Perplexity Comet to fill my Kroger cart is one example, but here’s another: Two months ago, I asked Comet to cancel my trial subscription to Clear airport concierge service. I actually signed up for Clear, and from there Comet navigated the labyrinthine menu system, initiated the cancellation request, and then handled the obligatory customer support chat on its own. What could have taken five or 10 minutes took seconds. (My wife’s AI-free cancellation experience went very differently.)
I’ve been frustrating AI agents beforebut I think there is potential when it’s built into the browser you’re already using and focused on reducing mindless busy work.
What are the trade-offs?
This is where I struggle with these AI browsers: most of the time, I don’t want to use them and feel anxious about doing so.
Worst of all, these browsers represent security minefields. It can include a web page that appears benign to humans Hidden instructions for artificial intelligence agentstricking them into stealing information from other sites – an attack method known as spot injection. “If you log into sensitive accounts like your bank or your email provider in your browser, just… Summarizing the Reddit postIt can result in an attacker being able to steal your money or private data wrote security researchers at Brave“Last week No one has figured out how to solve this problem.
If you can get past the security nightmares, the actual browsing features are subpar. Neither ChatGPT Atlas nor Perplexity Comet support vertical tabs — a necessary feature for me — and they don’t have a tab finder or way to search for recently closed pages. Atlas also does not support saving sites as web applications, selecting multiple tabs (for example, to close all at once with Cmd+W), or customizing the appearance. Compared to all the cool new AI features, the web browsing part can seem like an afterthought.
Regular web searching can also be difficult, although you may need it sometimes. When I typed “Sichuan Chili” into ChatGPT Atlas, a lengthy description of the Chinese chili appeared, not the nearby restaurant whose website and number I was looking for. (These browsers should be more like Dia, which intelligently routes queries to AI or Google based on what you ask for.)
Meanwhile, standard AI harassment is still implemented in the browser. Filling up my grocery cart felt like a victory, but on other occasions, the AI hit inexplicable walls and ended up wasting even more time.
There may be other costs to using these browsers as well. The AI still has limits to use, so this all eventually becomes a ruse to attract more people to the paid levels. beyond that, Atlas analyzes the pages you visit frequently– To build a “memory” of who you are and what you are. Don’t be surprised if this translates into deeply targeted advertising as OpenAI rolls out Find ways to monetize users for free.
How do I deal with it?
For now, I only use AI browsers in small doses when I think they can solve a particular problem. Until then, I won’t log into my email, bank accounts, or any other accounts where a security breach could be catastrophic. It’s too bad, because email and calendars are areas where AI agents can be really useful, but the security risks are very high (and Well documented).
However, AI browsers have an air of inevitability now, especially with OpenAI entering the scene. While some browser makers may reject AI on principle.As did VivaldiOthers like Chrome and Edge will continue to borrow ideas from the likes of Comet and Atlas, pulling them straight into their main browsers. The reality is that you may soon be using an AI-powered browser whether you intend to or not.
This story first appeared on advisorJared’s weekly newsletter about tech tips subscription Get more tips every Tuesday.
The early deadline for Fast Company’s World Changing Ideas Awards is Friday, November 14, at 11:59 PM PT. Apply today.
(tags for translation) Browser
ــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــ
