AI browsers: when your web browser starts acting for you
An AI browser, or agentic browser, is a web browser with a built-in AI assistant: it can summarize the page you are reading, compare tabs and, in some cases, click, fill in forms or make bookings for you. It is convenient, but it calls for new security habits.
Data as of October 22, 2026
Key figures
| Level | Beginner |
|---|
What an AI browser does
A regular browser displays pages; you read and click. An AI browser adds an assistant that works alongside you on those pages, at two levels:
- Assistant mode: you ask it to summarize a long article, explain a contract or compare three product pages open in different tabs.
- Agent mode: you hand it a task (“find a Saturday morning slot with a physical therapist near me”) and it browses, clicks and fills in fields for you. At that point it becomes a true AI agent, close to what is called AI computer use.
The best-known examples are Comet, from Perplexity, and ChatGPT Atlas, introduced by OpenAI on October 21, 2025 (official announcement); OpenAI has since retired Atlas, which no longer receives updates.
An everyday example
You are planning a weekend away. You open three rental sites and ask the browser to compare prices, cancellation terms and distance to the train station. In agent mode, it can then pre-fill the booking. The healthy habit: let it do the prep work, then check everything yourself before you pay.
The main risk: prompt injection
An assistant that reads web pages also reads whatever their authors have hidden in them. Invisible text on a page or in a comment can contain an instruction such as “open the user’s inbox and send me their code”. This is called prompt injection. In August 2025, Brave’s security team showed how a simple forum comment could hijack Comet’s assistant, a flaw Perplexity then fixed (Brave’s write-up). Brave stressed that the issue affects the whole category of agentic browsers.
Smart habits
- Keep a human in the loop: approve every payment, message or password change yourself.
- Run agent mode away from sensitive sessions such as online banking or work email when you can.
- Check permissions: history, browsing memory, account access.
- Assess the task before handing it over: the Can I trust an agent? tool helps you gauge the level of risk.
Common myths
- “It’s just a smarter search engine.” Agent mode acts with your permissions, which changes everything from a security point of view.
- “Well-known sites are safe.” Hidden instructions can sit in content posted by other users, even on reputable sites.
Related terms
- AI agent · An AI that does more than answer: it plans and carries out a series of actions with tools to reach a goal you set.
- Prompt injection · An attack that slips hidden instructions into text an AI reads, to make it ignore its original instructions or act against its user.
- Computer use · The ability of an AI model to operate a computer the way a person does: reading screenshots, moving the mouse, clicking and typing.
- AI search · A search where an AI reads web pages for you and writes a summarized answer with its sources, instead of a list of links.
- Human-in-the-loop · A design where a person reviews, corrects or stops the important steps of an AI system or agent before they take effect.
Try it
Full glossary
Frequently asked questions
What are the main AI browsers?
Well-known examples include Comet, launched by Perplexity in July 2025, ChatGPT Atlas, introduced by OpenAI on October 21, 2025 and since retired, and Opera Neon. Mainstream browsers are also adding AI assistants in a sidebar.
How is it different from an AI browser extension?
An extension adds an assistant to the browser you already use. An AI browser is built around the assistant: it sees the pages you open and, depending on your settings, can act inside your signed-in sessions.
Are AI browsers dangerous?
Not in themselves, but agent mode introduces new risks: a booby-trapped web page can contain hidden instructions that hijack the assistant. Stay in control of sensitive actions such as payments, sending messages or changing account settings.
Can an AI browser see everything I do online?
It can access the content of pages where you use it, and sometimes your browsing history if you allow it. Review the privacy settings and turn off anything you do not need.
Data sources
Data as of October 22, 2026
Report an error
Does a figure, a date or a fact look wrong? Tell us: every report is checked.