Firefox 148 Introduces AI Kill Switch for Total User Control

For years, browsers have been quietly stuffing themselves with generative AI, summaries here, copilots there, often without asking. Mozilla is now doing the opposite. With Firefox 148, the company is giving users a single, unmistakable choice: keep AI, customize it, or shut it down entirely.

The update, officially landing on February 24, 2026, introduces a new AI Controls hub inside Firefox settings. At its center is a blunt option labeled “Block AI enhancements.” Flip it on, and every generative AI feature in the browser disappears, instantly and permanently.

This isn’t a temporary pause. Once enabled, the setting survives future updates and blocks not only today’s AI features but anything Mozilla adds later. No pop-ups. No re-enabling. No “try this new AI tool” nudges.

A response to browser AI overload

Mozilla says the move is aimed squarely at users who feel overwhelmed by the way AI has become unavoidable in modern software. While other browsers continue to expand AI integrations by default, Mozilla is positioning Firefox as the opt-out alternative, a browser where AI exists only if you want it.

The change is already live in Firefox Nightly for testers, with the stable release coming later this month. For new installs, AI features remain off unless users explicitly enable them. Existing users keep their current settings untouched.

Control without the “nuclear option”

The master kill switch is only one part of the system. Firefox 148 also exposes granular toggles for five individual AI features, allowing users to disable them selectively instead of wiping everything at once:

• AI-powered translations, handled locally on the device
• PDF image alt-text generation, designed to improve screen-reader accessibility
• AI tab grouping, which suggests and names tab collections
• Smart link previews, offering summaries on hover
• The AI chatbot sidebar, which can connect to external services like ChatGPT, Gemini, Claude, Copilot, or Mistral

Each feature can be enabled independently, and none activate unless a user opts in.

Privacy-first by design

Mozilla is leaning heavily on local processing to support its privacy claims. Features like translation and PDF descriptions run on-device using compact models such as SmolLM2-360M, delivered via WebAssembly. In these cases, page content never leaves the user’s machine.

Cloud-based data sharing only occurs if a user intentionally connects a third-party chatbot through the sidebar. Even then, Mozilla says AI models are downloaded only when a feature is first enabled, not preloaded in the background.

Backing this stance with money, Mozilla is committing $1.4 billion from its reserves toward what it calls “trustworthy AI,” including transparency tooling and open research aimed at reducing dependence on closed, proprietary models.

A philosophical shift at the top

The update follows a clear message from Anthony Enzor-DeMeo, who took over as CEO in December 2025. His position is simple: AI should never be mandatory.

Internally, Mozilla is framing Firefox 148 as part of a broader effort to build a coalition, a self-described “rebel alliance”, of developers and researchers who want AI to remain decentralized, inspectable, and subordinate to user choice.

Why this matters

Firefox 148 isn’t rejecting AI. It’s redefining the power balance. At a time when browsers increasingly decide how much AI users get, Firefox is handing that decision back, with a single switch that actually means “off.”

In the current browser landscape, that may be the most radical feature of all.