AI browsers haven’t gone anywhere, and this one claims you can have personalization without giving up your privacy
If you’ve been following the AI space even remotely closely, you likely recall almost every AI lab had an AI browser phase. Perplexity and OpenAI both built standalone browsers with AI at their core, called Comet and Atlas respectively. Google, already owning the browser everyone else was trying to replace, went all-in on wiring Gemini straight into Chrome. The list didn’t end at the AI labs though, and browser companies got in too. Opera shipped Neon, and The Browser Company killed off its cult-favorite Arc to bet everything on an AI browser called Dia. We even saw Norton, the antivirus company, launch an AI browser called Neo. All that to say, the AI browser craze was very much in full swing. And while the hype surrounding most of these browsers seems to have faded, a new AI browser called Jatter launched recently claiming to respect your privacy. Of course, I had to take it for a spin.
Jatter is “the AI browser that respects your privacy”
The creepy part is that it isn’t creepy
Most AI browsers, including each one I listed above, now ship with agentic abilities front-and-center. Agentic essentially means the browser can do things for you, rather than just answering questions. You can ask them to book a flight for you, fill an entire form, add items to your cart, work through a multistep task while you go do something else, and so on. You hand it your intent, and it drives. Jatter, despite being positioned as an AI browser, does absolutely none of that. There’s no agent, and the browser just…reads. It’s built around the concept of personalization. It watches the sites you log into, learns from them, and hands back what it calls “personal answers” drawn from your own browsing. That’s the whole pitch, and it’s why it positions itself as privacy-first. The reasoning goes that if the browser never acts on your behalf, it never needs the keys to your accounts. Jatter leans hard into that framing, and its team makes a lot of promises. They claim no data, login information, or personal details are saved anywhere. Everything is encrypted on your device before it syncs, wrapped in end-to-end encryption so that even Jatter’s own servers supposedly can’t read it. Reportedly nothing is ever used to train a model, and whatever it does learn, you can wipe with a single click. That said, these are all just claims. To Jatter’s credit, the company behind it, previously known as Beacon, has spent years building privacy-branded tools. It’s previously built an encrypted location-sharing app and an anonymized chat app. Its founder has publicly framed his work as a stand against surveillance capitalism too. While you and I can’t really audit any of the big claims from the outside, the conviction does at least seem genuine. That said, sincerity isn’t proof, and there’s one thing you can check without a security audit: what the browser actually does once you begin using it.
Jatter is built around five apps
Grounded in your life
Jatter is available on macOS and Windows, with mobile apps on iOS and Android. It’s a full-fledged Chromium build, meaning it’s built on the same open-source foundation as Chrome, Edge, Brave, Opera, and most other browsers you can name. The setup is fairly simple. You install the browser, sign in, and verify your email. From there, you’ll be greeted with Jatter’s version of the new-tab page. I’m the kind of person who is very influenced by the aesthetics and UI of the tool I’m using, and while Jatter has an incredibly cute logo, the UI is far too bland for my liking.You’ll find the address bar up top that you’re used to (ironically, with Google’s AI Mode prompt), and five apps laid out in the center: Chats, Personal Answers, Notes, Maps, and Settings. Below them sits a single Ask Jatter box, which is where most of the actual back-and-forth happens. That said, Chats is where you’ll find all your previous conversations with Jatter. Personal Answers is the panel that actually shows you what Jatter has learned, organized by site and topic, and lets you query it directly. Notes and Maps are the two apps Jatter describes as AI-native. Beyond Notes being a place where you can jot things down, Jatter says it learns from whatever you put there. For instance, Jatter says it can take the meeting summaries, research, and half-formed ideas you drop into Notes and fold them into its answers — suggesting an action plan after a client call, say, or synthesizing something usable out of a page of brainstorming. Maps works on the same principle, but for places: it pulls in your travel plans, saved locations, and past searches to offer trip recommendations and location-specific advice, with interactive maps built into the browsing experience. In practice, both apps are really just more intake pipes into the same personalization engine.
The whole point of Jatter is Personal Answers
Notes, Maps, and a browser that pays attention
Whenever I test a new AI browser, I like to use it exclusively for a few days and put it through some basic testing. Now, the Chat tab within Jatter is no different than your typical AI chat interface at first glance. You ask it a question, it searches the web or uses its training data, and gives you an answer, complete with sources if it searched the web. Ask it to explain something, draft an email, summarize an article, and it behaves exactly like the chatbot you already have open in another tab. On that level, there’s nothing to write about. The reason Jatter, exists, though, is that it’s meant to answer questions about you. AI chatbots like Claude and ChatGPT do this by building up memory about you based on chats you have within them. Over time, they pick up that you’re a journalist, or that you’re planning a trip, and lean on that context in later conversations. But that memory is bounded — it only knows what you’ve actually told it, inside the chat window. Jatter’s whole premise is that this is far too narrow. Why should your AI only know the version of you that shows up in a chatbox, when the fuller picture (your bookings, your searches, your bank portal, your inbox) is sitting right there behind logins it could just read?
Related
You don’t need Perplexity’s Comet browser; just use Playwright instead
Powerful automation without privacy nightmares
That’s the leap Jatter makes. Instead of waiting for you to tell it things, it learns from the sites you use, and answers questions grounded in your actual digital life. So to test it, I did the obvious thing: I gave it something real. I was, at the time, deep in planning a trip to Malaysia. So I logged into the sites I’d been using to plan it, switched on personal answers, and started asking Jatter about my own trip. For instance, I logged into Booking.com, where I’d sorted out my accommodation, and the moment I did, Jatter surfaced a small prompt asking whether I wanted to enable personal answers for the site. Right away, I like the fact that consent is opt-in, and it’s per-site. You decide what Jatter is allowed to read, and you do it one site at a time, with a plain “Never” or “Enable” rather than a single blanket switch that swallows your whole browsing life at once.
Once I had granted it permission, I went to Jatter’s chat interface and asked it which hotels I had booked for the trip. At this point, Jatter wasn’t connected to my Gmail, my calendar, or anything else. I had enabled it on exactly one site, Booking.com, and I had made the reservations a week earlier on another browser. So far, so good! I then did some browsing related to my trip. I looked up some blogs, read a few TipAdvisor pages, browsed some Klook activities, went down a Reddit rabbit hole, and ran the usual scattered Google searches you fire off when you’re half-planning, half-daydreaming about a trip. Then, I went back to chat and started asking Jatter about questions related to all that browsing. I started with something small and a little silly. I’d seen something about MrBeast’s chocolate being sold in Langkawi and had made a mental note to look for it, so I asked Jatter what chocolate I wanted to get there. It came back empty, and said it couldn’t find anything about chocolate, only my upcoming trips to Kuala Lumpur and Pantai Cenang. Fair enough. But I knew I’d searched for it, so I pushed: did I search for something about chocolate? This time, it noted that I had searched for “langkawi mr beast chocolate” on Google. I found it interesting that a LLM couldn’t connect the dots between the two, which sort of defeats the purpose entirely. Regardless, then I decided to ask Jatter what it knew about my trip, and it laid out a list of places it had connected me to: Pantai Cenang, Kuala Lumpur, Ipoh, Singapore, Bangkok, Bukit Bintang, Dang Wangi, plus both my current hotels. Some of those I hadn’t booked anything in. Some I’d only glanced at in a search result. It told me my Google and TripAdvisor activity pointed to an interest in “things to do in Langkawi,” that I’d been on Klook and Booking.com looking at tours and accommodation. To be fair, none of the above is hidden. Jatter uses your browsing history, says so, and even gives you a panel called “Browsing History” where it’s all laid out. What’s interesting here is the consent I’d given was per-site. I only enabled one site, Booking.com, yet the reach clearly wasn’t per-site. You can also delete what it learns. There’s a “Delete learned history” button right in the panel. I hit it, confirmed, and asked Jatter the same questions again.
This time, it drew a blank on the chocolate search, which was the one thing that lived only in my history.
When I asked it about my hotels again, it answered the question correctly because it just re-read Booking.com on the spot.
So deletion genuinely clears what it’s stored. However, it just doesn’t stop Jatter from re-learning the moment you ask about a site you’re still logged into.
Nothing here you can’t get safer elsewhere
While Jatter does what it promises to do well, I’m not entirely sure it does anything you can’t get somewhere else. Yes, it claims nothing is stored, everything’s encrypted, and even its own servers can’t read your data. But those are promises I can’t verify, and every serious browser in this space now makes some version of them.
What I can compare is how each one handles the tension at the heart of all these tools: an AI is only useful if it knows things about you, and the things it knows are exactly the things you’d least want leaking. And on that front, Jatter is oddly middle-of-the-pack for something selling privacy as its headline feature. Dia keeps its history-based context to the last seven days, stored locally. Norton Neo, underwhelming as it was to actually use, was built by an actual security company to keep your data local by default and to defend against the hidden-instruction attacks these browsers are uniquely prone to. Nonetheless, I’ll be keeping an eye on Jatter. The idea underneath it is a good one, and it managed to impress me more than once during my week using it.
Diterbitkan : 2026-07-10 10:00:00
sumber : www.xda-developers.com
