Mozilla announced that the browser Firefox is "fully AI-based", but the development community led by Waterfox immediately turned against it
After the Mozilla Foundation changed its CEO, Firefox announced its full embrace of large-scale language models, triggering Waterfox's counterattack and privacy concerns. The browser market has an obvious rift between the AI and de-AI poles.
(Previous summary: After accepting Dogecoin was scolded, the Mozilla Foundation: will reject donations from PoW cryptocurrency)
(Background supplement: Vibe Coding creates workplace competitiveness: free resources and tools for getting started in AI engineering)
Contents of this article
Decades ago, the Firefox browser became popular around the world Mozilla, after taking over as its new CEO, announced the "AI-first" route, planning to put large language models such as ChatGPT and Claude directly into the core of Firefox. This policy, personally exposed by new CEO Anthony Enzor-DeMeo, aims to reverse the dual dilemma of falling below 3% market share and over-reliance on Google search profits.
As soon as the news came out, Alex Kontos, who maintains the long-term fork project Waterfox, immediately expressed his opposition in a public response, and the browser industry entered a debate on the "soul position".
Firefox is trying to "change its soul" for revenue
Firefox once used privacy and autonomy as selling points, but after the profit sharing from advertising and search shrank, operating cash flow became increasingly tight. New CEO Enzor-DeMeo has put his chips on AI, such as placing summaries, rewrites, and search assistants next to tabs, and retaining a "global close" button to appease the community.
This may seem like a compromise, but it actually locks the AI model into the bottom layer of the browser, which subverts Firefoxâs original product positioning for community support.
Waterfox: Browsers should not become secondary proxies
The Waterfox team believes that AI plug-ins and native tools can be used, but turning LLM into a default component will allow the browser to automatically filter information and rewrite content, making it difficult for users to detect the boundary between real web pages and AI-generated results.
Alex Kontos emphasizes:
We must distinguish between auditable machine learning and unpredictable LLM. The browser should be your tool, not your broker.
Waterfox retains the Gecko core and supports Widevine DRM, but explicitly refuses to integrate LLM, giving hardcore users who value digital sovereignty a retreat.
Building LLM into the rendering layer is equivalent to inserting a potential leakage channel into each page. The community stated that prompt word injection allows attackers to modify AI behavior through hidden instructions and further read passwords or leak cookies. To improve the quality of the summary, the browser also needs to transfer the page context back to the cloud and transfer it between third parties such as OpenAI and Anthropic, making the transmission chain increasingly opaque.
For users who have been accustomed to using the "Go Tracking" extension function for a long time, being able to turn it off does not mean that they can be trusted.
The final user migration of browsers
The browser fronts are rapidly dividing: Arc, Edge and the new Firefox stand in the "one-stop AI entrance" camp, while Waterfox, LibreWolf, etc. hold high the flag of local computing and minimal permissions.
The former competes for mainstream traffic, while the latter defends a small number of privacy believers with high stickiness. Whether AI can save Firefox, the final results will be seen in the active device and search profit sharing reports in mid-2026.
Seriously, if you care about the capabilities of current AI tools, why would you use Firefox? The disbandment of the Mozilla Japanese community is not without reason. Mozilla deeply hurt the hearts of old users.
