Head of Firefox: Control over AI and a different web is possible

Mozilla recently made executive leadership changes. Ajit Varma is new to his role as Head of Firefox. We spoke with him about the future of browsers in the age of AI, the importance of trust, and where Firefox intends to position itself.

heise online: Firefox is growing, at least the mobile version of the browser. And there has been so much change within Mozilla. But actually, there are more browsers out there than ever before. Where do you position Firefox?

Ajit Varma: It's been an interesting year, because a year or two ago, no one really talked about browsers. And that made browsers low consideration, people just typically use whatever came with their operating system. So, the good thing about the conversation is now people are actually talking about what browser they should use, and that means that people are considering it, which is really beneficial for Firefox.

The thing that really distinguishes our approach versus many of the other browsers is that we do not build our own LLMs that we then want to promote. We are working with a number of LLM providers that we integrate, but don’t push a specific LLM. When you look at some of the newer browsers that have emerged, or even existing browsers, there is a big push for deep integration of AI. Edge looks like a CoPilot app. Gemini is getting deeply integrated. And all of these browsers only allow you to use the AI that their company has built. Whereas with Firefox, we are very focused on choice, just like you can pick any search engine, you can pick multiple AIs. We're going to give users the choice.

But then, ultimately, if you're a more sophisticated user and you want to bring your own AI and build your own AI, we'll actually let you plug that into the browser for certain functions as well. This is a big differentiation for us, and I think that it gives a lot of opportunities to the non-big, giant AI players to actually participate in browsers as well.

heise online: There are, on the one hand, a lot of people talking about AI all the time, but on the other hand, there are a lot of people annoyed, too. And you probably know it best because they started to protest when you introduced AI to Firefox. Why do people not like AI, or at least don't want it in Firefox?

Ajit Varma: I think that there are a few buckets of concerns with AI. I think a lot of the Firefox users are concerned about the societal impacts of AI – from environmental concerns to privacy to job loss.

We want to make it really clear to users, that if the feature is powered by AI people can pick and choose what they want to use, and so a user could say, translations are valuable, but they don't want to do summarization.

Or they can use what we've recently launched, which is AI Controls. This is a single switch that allows a user to turn off all of AI. People are wondering if turning on AI controls means that we're going to force users to use AI, and that's definitely not the case. Everything is by choice, and so we try to make it transparent before you engage, there's a screen that says: Do you want to use this? And then we try to do the maximum privacy-preserving methods as well. For things like translations, we download to local models, we don't share anything with the cloud.

heise online: In Germany and in the EU, especially politicians, are talking a lot about responsible AI. But I often miss content, like, what is responsible AI, or is there a chance for responsible AI? What do you think about that?

Ajit Varma: I think that there are definitely many different approaches that AI companies are taking. At one extreme, you have companies that are saying there's no safety, generation of any kind of content is what everybody wants. And then there are open-source models that are very transparent in what they're creating and what they're doing. And you have other models that are local. But all these things have different trade-offs: in terms of the level of capabilities, the level of interest, maybe to users.
If everyone is using a particular browser, and that browser company has an AI that they want to push, and that browser AI makes a decision that this is the way they believe the world should look; it's really detrimental to people living the lives that they want. By offering the ability to bring whatever AI you want, or no AI at all, hopefully people can make the right choice for themselves.

heise online: So you won't start to build your own LLMs?

Ajit Varma: We have no plans to. We are looking at open-source models and talking to other companies, but we are not developing any models in-house.

heise online: I think AI is like a race, and currently, everybody wants to win, and winning means winning everything. Do you think there even is a chance to win, or is it more like everybody has to be on board because nobody knows what happens next?

Ajit Varma: I think that there are a lot of scenarios that could happen. On the negative side LLMs are very expensive to build, foundational models cost a lot of money. There's just very few companies that can afford it. Even when you look at companies like OpenAI – they're losing money on every query right now, and they can't figure out how to monetize. The Googles, the Microsofts, the Metas, they're going to have control. And their motivations are very similar. They are looking at LLMs and AI as a way to sell more ads. They're ad-driven businesses. And then you get into the question, if the answer AI is giving you, is the best answer or the best to make the most money? These companies are probably going to optimize for shareholders.

The key is to reduce the cost to build LLMs. I think that there are already dynamics that we are seeing. If you look at Chinese models, they are built much more cost efficiently. They're doing distillation of other models, and all the companies who own those models are complaining - but those companies are also using the open web in order to generate their content, too.

heise online: Sometimes I hope AI will really change the internet and the whole problem with monetization. Do you think it is possible that the internet will change and there will be a different kind of monetization?

Ajit Varma: We have seen it happen before – 15 years ago, everything was just a web, and nobody was paying for anything. Especially, this is true for music. People would listen to music for free, they would use file-sharing networks in order to download. But then, companies like Spotify and Apple Music created an easy way to consume it. You got to the point where you would rather pay $5, $10 a month because it's just more convenient, a better experience, and it's legal.
You might see the scenario where I just pay a few pennies for every time I read an article, whether that's in AI or however I'm getting the content. That was never really the value proposition that happened when search engines came. And this was actually a big question 20 years ago. Fox sued Google, saying you can't look at our content, but they didn't have a choice, because if they removed their content from Google, they would lose so much traffic that they had to say, we'll give you our content for free.

If companies had said no, you have to pay for this content because you're showing ads on the value of this content, and you have to share that revenue. It would be a very different world today. You'd have a lot more journalism and a lot less of the social media, doomscrolling and bubbles that are happening right now. I am hoping with AI there are these models that are definitely possible if people, or companies are promoting it. But it doesn't look like it's going down that way, unfortunately.

heise online: The big AI companies say that the whole way we communicate with a computer will change, too. Satya Nadella once said it will change the computer like the mouse did once, and others are betting on hardware devices like glasses or pins. What do you think is the role of a browser in the future, then?

Ajit Varma: If you look at the term “browser”, it very much implies you're clicking on links and you're browsing sites. It's not interactive, it's not two-way and that's the historic web page developed 20 years ago But the power of a browser is to actually be a tool for how you engage with open content and get to any company without any gatekeepers there.

If you look at ChatGPT or Facebook or YouTube or even Google; they were possible because you could access them without a gatekeeper in place. If the gatekeeper is a company that is competing with you, I can guarantee you that there will not be the opportunity for competitors and competition to emerge. And less competition often means worse products for users.

With AI, you can point to content and say: How can you edit this? How can you transform this page? How can you make it satisfy my needs? You can look at things like summarization. I can look at content and say, remove political bias from content, or give me an alternative viewpoint. Instead of me having to get up every day and look for jobs, I can create bots to help me accomplish the tasks I want to do. And so this is where the browser goes from just browsing to how do you help someone accomplish the journey they're trying to accomplish.

I think there will be different surfaces for this, just like how mobile phones emerged and people started using apps instead of using a browser. And I think that you might see voice-activated devices, you might see glasses. But I think that if that's all there is, then you see a lot of the things that enable that to be possible, like all the content that's created on the web. If the economic model doesn't exist, the thing that made all the AI possible, which is all the data to train on, doesn't exist, then it gets worse.

heise online: Do you think, in the future, in the coming years, everything will be AI-driven?

Ajit Varma: Change is certainly accelerating faster than I expected it. Looking specifically at software tools, I think it's very ironic that one of the industries that's probably going to be disrupted the most and the earliest by AI creation is the jobs of engineers and software engineers. The tools that exist now versus 6 months ago – it is just incredibly better.

heise online: I think it's pretty difficult for many people to be on board with this whole change, and even for me, I sometimes just want to use my browser and not AI, I know how the web works for me. Do you think this will be a problem for many people – or more of a chance for browser providers?

Ajit Varma: We definitely recognize that. Many Firefox users are very vocal that they don't want AI now or ever. We actually paused releasing AI features in order to launch AI controls. The last 3 months, we've reallocated our teams to really make sure that this gets out first before anything new comes out.

We do look at all the forums, we do look at all our feedback from customers, and if our customers want no AI at all, then that's the path we'll go down. If they want choice, that's the path we'll go down. And right now, we're looking at choice as what we think people want.

heise online: What about trust? Is maybe trust something we can rely on, even more in the future?

Ajit Varma: One of the things that we are leaning heavily into is becoming the most trusted software company and trust has different dimensions. One of those is being clear on our intent, we're transparent in what we want and that we give users choice. Within Firefox we built a lot of things for trust, we built containers that really segregate all the data and make sure that nothing is shared. We built our sync in a way that it's end-to-end encrypted. But we never did a good job of explaining this to users. We asked users: Is it hard to log in to Firefox? They say yes. We require passcodes, we require passkeys, and that's to make sure that it's all encrypted data that no one has access to. And so the next couple of months, we're going to try to highlight more of those features, but then also launch new features, like an integrated VPN function that routes web traffic through a proxy server and allows users to conceal their IP address.

And the other area is: Firefox is owned by a non-profit and we're not trying to optimize for shareholder value; we're optimizing for creating the best browser to create the best user value. And this means that we are balancing features that can generate sustainable revenue and features that create a better user experience.

heise online: Thank you so much...

Ajit Varma: Germany is definitely one of the most important markets for Firefox, and we've seen growth in mobile, and I think over the next year we are definitely prioritizing a lot of the users that we see in Germany and the needs that we see there. I am so excited to continue to get feedback on how we can make the best browser.

(emw)