After Mozilla Foundation replaced its CEO, Firefox announced a full embrace of large language models, triggering a backlash from Waterfox and raising privacy concerns, leading to a noticeable rift in the browser market between AI and anti-AI. (Previously: After being criticized for accepting Dogecoin, the Mozilla Foundation: will refuse donations from PoW cryptocurrencies) (Background Supplement: Vibe Coding Builds Workplace Competitiveness: Free Resources & Tools for AI Engineering Beginners)

Table of Contents

• Firefox attempts to “change its soul” for revenue
• Waterfox: Browsers should not become secondary proxies
• Last User Migration of the Browser

Mozilla, which became globally popular due to the Firefox browser decades ago, announced an “AI-first” strategy after the appointment of a new CEO, planning to integrate large language models like ChatGPT and Claude directly into the Firefox core. This policy was personally revealed by the new CEO Anthony Enzor-DeMeo, aiming to reverse the dual dilemma of market share dropping below 3% and excessive dependence on Google for search revenue.

Once the news broke, Alex Kontos from the long-term maintenance fork project Waterfox immediately expressed opposition in a public response, leading to a debate in the browser industry over the “soul position.”

Firefox attempts to “transform” for revenue.

Firefox once marketed itself with privacy and autonomy as selling points, but as advertising and search revenue dwindled, its operating cash flow became increasingly strained. The new CEO Enzor-DeMeo has bet on AI, such as integrating summaries, rewriting, and search assistants beside tabs, while retaining a “global close” button to reassure the community.

This seems like a compromise, but in reality, it locks the AI model into the underlying layer of the browser, which subverts Firefox's original product positioning regarding community support.

Waterfox: Browsers Should Not Be Secondary Proxies

The Waterfox team believes that AI plugins and native tools can be selected, but making LLM a default component would cause the browser to automatically filter information and rewrite content, making it difficult for users to perceive the boundary between real web pages and AI-generated results.

Alex Kontos emphasized:

We must distinguish between auditable machine learning and unpredictable LLMs. The browser should be your tool, not your broker.

Waterfox retains the Gecko core, supports Widevine DRM, but explicitly rejects the integration of LLM, providing a way out for hardcore users who value digital sovereignty.

Integrating LLM into the rendering layer is equivalent to inserting potential leakage channels into each tab. The community indicates that prompt injection allows attackers to alter AI behavior through hidden commands, further gaining access to leaked passwords or cookies. To improve summary quality, the browser also needs to send the page context back to the cloud, circulating among third parties like OpenAI and Anthropic, making the transmission chain increasingly opaque.

For users who are accustomed to using the “track” extension for a long time, turning it off does not mean it can be trusted.

Last User Migration of Browser

The browser front is rapidly diversifying: Arc, Edge, and the new Firefox stand in the “one-stop AI portal” camp, while Waterfox, LibreWolf, and others proudly wave the banner of local computing and minimum permissions.

The former competes for mainstream traffic, while the latter holds onto a small but highly engaged user base of privacy advocates. Whether AI can save Firefox will be evident in the active devices and search revenue sharing report in mid-2026.

To be honest, if you care about the current capabilities of AI tools, why would you use Firefox? The dissolution of the Mozilla Japan community is not without reason; Mozilla has deeply hurt the hearts of long-time users.