Gaining a foothold in the web browser market is tough. There are some long-established players unwilling to give up their lead and with the resources to figure out how to hang on to it.
But times change. While Microsoft once sent Netscape packing, bundling its Internet Explorer browser with Windows – a move that ultimately meant it fell foul of competition laws in the European Union – the company was subsequently overtaken by Google’s Chrome browser before it retired Explorer in favour of Edge.
If you thought that meant it was all settled and that the winners had been chosen, you would be wrong. The tech world is always on the hunt for the next big thing, and the potential for AI in your web browser is fuelling a whole new browser war.
Artificial intelligence (AI) browsers are a rapidly evolving model. But they are more than just summarising web pages and answering a few questions. AI browsers include advanced features that let the technology take the wheel and carry out tasks independently.
READ MORE
If you want to try out a few, there are options out there.
ChatGPT Atlas
OpenAI wants to be all things to all people. That can be an AI-fuelled assistant or a tutor, a video generator or a research assistant. In the future, it could also include a social media app, or shopping through ChatGPT.
For now, it has set its sights on your browser. There were suggestions that it may have its sights on Google Chrome, should the tech giant be forced to sell the popular browser to placate competition authorities.
Instead, OpenAI launched Atlas, an AI browser that is based on chromium, the underlying code for Google’s Chrome browser.
As you might expect, ChatGPT is built into the heart of the browser, alongside an agent that will independently run tasks for you.
The chat it runs in a sidebar allows you to ask questions about the webpage you are looking at or edit text.
Your data is important too. Atlas draws on past conversations you have had with ChatGPT to inform its future interactions with you, and also pulls in data from your browser history – an optional feature.
Many of the advanced features of the browser are reserved for OpenAI’s paying customers. Although if you have a subscription with the service, you can use the AI agent to book hotels, create documents on your behalf or go on a shopping spree.
For now, Atlas has only been launched for Mac. That may seem like an unusual move, given its close association with Microsoft. But Sam Altman’s company is notching up the deals. Already, it has a partnership with Apple to broaden out its Apple Intelligence for general requests that Apple can’t answer based on your own data.
Perplexity Comet
If you aren’t familiar with Perplexity, you may soon be. The AI search company has been winning people over with its app that offers an AI-powered search engine that summarises information for users. It includes academic papers, a section for searching SEC filings and a social option for searching discussions and opinions.
In July, it branched out in the browser market with Comet. Initially available only on desktop, Comet is designed to take meaningful actions on your behalf. That can be anything from checking and summarising your email for you to booking meetings on your behalf or compiling your schedule for the day.
You can use it to build basic websites too, with just a few natural language instructions.
It can also search your browsing activity for you, and use it to create actions. For example, it can book a restaurant whose details you searched for online on Friday.
It is now available on Android too, with much of the same functionality.
Expect that to expand in the coming weeks, with plans to include a conversational agent that can search sites on your behalf.
The company isn’t leaving Apple users out. Although it hasn’t launched its iOS app just yet, it is in the works. However, Android pipped it to the post, possibly.
Google Chrome with Gemini
Google hasn’t added a whole new browser to its line-up. Instead, it has added Gemini to its Chrome browser as an extension, bringing the power of its AI to users.
Using the extension, you can find answers based on the tabs that are open on your browser, bringing important context to the conversation. You can also use Gemini to explain tricky concepts to you more easily, or act as a research assistant to help you dive deeper into a concept. There is no need to switch tabs, and Gemini will work with voice and typing inputs.
It keeps things familiar and is easy to access, but it only works when you ask it to – in other words, it isn’t constantly monitoring your activity.
Google is rolling out the feature gradually, starting with US users, and it is also coming to mobile. The Android version will work with anything on your screen, and there are plans to build Gemini into the Chrome app on iOS.
Opera One
If you call yourself the future of browsing, you have to aim big. Opera One gives you access to its Opera AI (formerly Aria) assistant free of charge, without needing an account.
It can generate images, answer questions and read you a bedtime story. Like other browser AI, it can access your web pages on command and answer questions based on what is on screen. And it not only generates images, it can also understand and interpret them.
But back to aiming big. Opera One offers access to not only its own built-in assistant, but also access to third-party models, such as Meta’s Llama, Google’s lightweight Gemma model, and even ChatGPT.
So you don’t need to lock yourself into one AI provider.
Edge with Copilot
If you have been actively avoiding ChatGPT but are using Copilot within Microsoft’s suite of products, we have news for you. Copilot is based on ChatGPT’s technology.
But the key thing is in how data is managed. While Microsoft’s AI may be powered by ChatGPT, enterprises can set limits on how their data is used. That is also true of the Copilot feature for Edge.
Copilot in Edge brings some of the AI expertise to the web browser. Like Gemini in Chrome, you can research topics, ask questions about your on-screen content, summarise web pages and videos, and even generate images in the sidebar.
Fellou
You may not have heard of Fellou, but you may well do in the future. Chromium-based Fellou offers basic browsing for free, but its advanced options – the stuff it really excels at – require credits to run. That includes agentic tasks, memory and so on.
But it can do a lot. You can use it to automate market research, design a curriculum for topics or even search for a job on your behalf.
The case against
As with all new developments, though, the technology inevitably has some downsides. There are many people who are wary of AI browsers, for several reasons. First, there is the idea of handing over so much personal data and habits to an AI company.
Malicious prompt injection has emerged as one of the top fears. This is where bad actors try to manipulate the AI processes by slipping hidden instructions into the content the AI assistants are processing. That can be completely unbeknown to users, with the assistants effectively working against them.
Perplexity has acknowledged the difficulties the new technology brings, saying its action-oriented design makes the browser more useful but also brings new threats.
“AI introduces vulnerabilities that were previously not possible with classical application security. And for the first time in decades, we’re seeing new and novel attack vectors that can come from anywhere,” Perplexity’s security team wrote in a blog post on an attack called prompt injection.
“It demands rethinking security from the ground up.”
Describing the security threat as an “unsolved problem across the industry”, the company says it has created some defences specifically for these use cases, implementing machine learning that is trained to spot hidden instructions before they run.
Among the threats it can detect are invisible text instructions such as zero font size or CSS instructions, text encoded into images that cannot be seen by the human eye, and goal hijacking, which can override the original query or compromise system prompts.
There are additional layers of security designed to weed out any attacks that slip through, such as requiring user confirmation for sensitive actions, such as placing shopping orders.
“Experience has taught us that security isn’t a feature to bolt on after launch, but a foundational requirement that requires reimagining how malicious action is conceived and where those attacks will come from,” the company said.
It isn’t the only one warning about the potential side effects of the new generation of browsers. Web company Brave has been researching the potential security vulnerabilities of the new technology, saying indirect prompt injection is a “systemic challenge” facing the entire category.
“Long-standing Web security assumptions break when AI agents act on behalf of users,” the company wrote in a recent blog post.
It has disclosed several vulnerabilities in these browsers, including Perplexity’s Comet, Fellou and Opera Neon, to the makers of the technology before publishing them more widely.
While the new technology could bring sweeping changes, as usual, it is not without its drawbacks – something users should keep in mind when they are weighing up the options.
