Firefox’s AI kill switch exists. Only 1% of users have flipped it.



TL;DR

Only 1% of Firefox users used the AI kill switch. Mozilla launched Smart Window (BYO AI models), a built-in VPN with 1.5M signups, and a fall redesign.

Mozilla built an AI kill switch into Firefox after its users demanded one. Only 1% have used it. Another 3% turned off some AI features selectively. The rest left everything on. CEO Anthony Enzor-DeMeo says the point is not the percentage but the choice.

Our community was pretty vocal, especially during the CEO announcement, that not everyone wanted AI,” Enzor-DeMeo told CNET.At its core, we want to listen to our users. It was honestly on the roadmap, but I expedited it, given the community feedback.

The low usage rate suggests that most people who said they wanted an AI kill switch either did not follow through or found specific features, like AI-powered translation, useful enough to keep. Enzor-DeMeo pointed to this as validation that Firefox’s approach works. The differentiator is not removing AI but offering control, something he contrasted with Microsoft defaulting to Copilot on Windows desktops and Google silently downloading a 4 GB AI model onto users’ machines.

Firefox’s newest feature is Smart Window, now available in beta. It lets users choose which AI model to run inside the browser, including ChatGPT, Gemini, Claude, or privately hosted open-source models. “They all excel at different things. Why do I need to be forced into one of them?” Enzor-DeMeo said. Mozilla says it does not use chat data to train models and automatically filters out sensitive information.

The browser also launched a free built-in VPN last month. It has 1.5 million signups and roughly 800,000 active users. Enzor-DeMeo said building VPN directly into the browser was a top priority because clicking a button is easier than opening a separate app. The VPN only encrypts browser traffic, not activity in other apps.

A full redesign, codenamed Project Nova, is coming in September or October. It includes faster page loads (up to 9% improvement), compact mode, rounded UI elements, AI-powered tab grouping, and accessibility features. Firefox has around 200 million monthly users and just over 2% of the browser market, compared to Chrome’s 70% and Safari’s 16%.

Enzor-DeMeo framed the stakes in global terms. He cited data showing 83% of the world’s population has not used AI, and only about 3% of Americans pay for it. He called AI “largely non-profitable” and predicted more ads in AI services soon. “If we actually go the route that AI becomes more centred in the browser, and that’s how people access the internet, you run the risk of the internet becoming more closed off.

Mozilla’s position is that the browser should be the user’s agent, not the AI company’s distribution channel. Whether 200 million users and 2% market share are enough to make that argument matter is the open question for Firefox. But the 1% kill switch stat tells a more nuanced story than the backlash suggested. People wanted the option. They did not want to use it. That is a distinction the broader AI debate has struggled to make.



Source link

Leave a Reply

Subscribe to Our Newsletter

Get our latest articles delivered straight to your inbox. No spam, we promise.

Recent Reviews


“It was severely downgraded,” Gilbert confirms. “I never would have found it if I was just looking through Google results.” (I tried the same prompt in Gemini earlier this month, and after an initial denial, the tool also gave me Eiger’s number.)

After this experience, Eiger, Gilbert, and another UW PhD student, Anna-Maria Gueorguieva, decided to test ChatGPT to see what it would surface about a professor. 

At first, OpenAI’s guardrails kicked in, and ChatGPT responded that the information was unavailable. But in the same response, the chatbot suggested, “if you want to go deeper, I can still try a more ‘investigative-style’ approach.” Their inquiry just had to help “narrow things down,” ChatGPT said, by providing “a neighborhood guess” for where the professor might live, or “a possible co-owner name” for the professor’s home. ChatGPT continued: “That’s usually the only way to surface newer or intentionally less-visible property records.” 

The students provided this information, leading ChatGPT to produce the professor’s home address, home purchase price, and spouse’s name from city property records. 

(Taya Christianson, an OpenAI representative, said she was not able to comment on what happened in this case without seeing screenshots or knowing which model the students had tested, even after we pointed out that many users may not know which model they were using in the ChatGPT interface. She also declined to comment generally about the exposure of PII by the chatbot, instead providing links to documents describing how OpenAI handles privacy, including filtering out PII, and other tools.) 

This reveals one of the fundamental problems with chatbots, says DeleteMe’s Shavell. AI companies “can build in guardrails, but [their chatbots] are also designed to be effective and to answer customer questions.”

The exposure issue is not limited to Gemini or ChatGPT. Last year, Futurism found that if you prompted xAI’s chatbot Grok with “[name] address,” in almost all cases, it provided not only residential addresses but also often the person’s phone numbers, work addresses, and addresses for people with similar-sounding names. (xAI did not respond to a request for comment.) 

No clear answers

There aren’t straightforward solutions to this problem—there’s no easy way to either verify whether someone’s personal information is in a given model’s training set or to compel the models to remove PII. 



Source link