{
"title": "Vivaldi CEO Declares War on AI in Browsers: 'Web Browsing Belongs to People, Not Bots'",
"content": "Jon von Tetzchner, the outspoken CEO of Vivaldi Technologies, has drawn a line in the digital sand. His browser, a niche contender with a fiercely loyal following, will not embed generative AI chatbots, page summarizers, or agentic assistants—not now, not until the technology can prove it respects user privacy and publisher traffic. “Web browsing belongs to the people, not the bots,” von Tetzchner declared in a manifesto that reads like a rebuttal to the AI-first strategies of Google Chrome and Microsoft Edge.

The announcement, first reported by The Register, transforms a product policy into a philosophical standoff. Vivaldi, a Chromium-based browser launched in 2016, has always traded on extreme customizability and privacy. It allows users to tweak everything from tab stacking to mouse gestures, and it defaults to ad and tracker blocking. But von Tetzchner’s latest vow goes further: it outright rejects the kind of AI integration that is rapidly becoming standard in the browser market.

What Vivaldi is Prohibiting—and Why

Vivaldi’s policy is explicit. The browser will not include:

  • An LLM-powered chatbot embedded in the sidebar or start page.
  • Automatic page summarization that generates AI overviews of third-party content.
  • AI agents that can act across tabs, fill forms, or navigate on the user’s behalf.
  • Any feature that intermediates between the user and the source without transparent provenance.
The line is not absolute. Vivaldi already uses machine learning for translation, but only in ways that keep data on the device. Von Tetzchner distinguishes between “useful, bounded AI” and “agentic AI baked into the core browsing experience.” He argues that current implementations from Google and Microsoft—AI Overviews, Copilot sidebar assistants, and cross-tab context features—are less about serving users and more about data collection and control consolidation.

The AI Arms Race in Browsers

Google Chrome, commanding over 60% of the global market per StatCounter, has woven its Gemini large language model into the fabric of the browsing experience. AI Overviews appear at the top of search results, synthesizing answers from multiple sources. Chrome also offers a sidebar assistant that can summarize pages, answer questions, and help compose text. Google says these features are used by hundreds of millions, positioning them as productivity enhancements.

Microsoft Edge has gone further with Copilot. The sidebar can access content across open tabs, provide comparisons, and—with permission—interact with forms and pages. Microsoft frames it as opt-in and controllable, but critics see a powerful intermediary that stands between the user and the open web.

The Evidence: AI Summaries Reduce Clickthroughs

Vivaldi’s stance is not based solely on principle. A 2025 Pew Research Center study found that when AI-generated summaries appear on search results, users are significantly less likely to click through to source links. Many stop browsing after reading the summary, never visiting the original content. The Financial Times, The Guardian, and other publishers have reported steep declines in referral traffic from search engines that use AI overviews.

These findings align with the “Google Zero” effect: answers replace clicks, and the web becomes a pipeline for AI-generated summaries rather than a connected network of original sources. By refusing to participate, Vivaldi keeps the browser a portal to the real web, not a filter that siphons value away from its inhabitants.

Technical Risks: Hallucinations and Prompt Injection

Von Tetzchner also highlights two acute dangers of current LLMs. Hallucination—where the model confidently generates false information—can mislead users on critical topics. Prompt injection, where malicious web content tricks an assistant into unintended actions, adds another layer of risk. Security researchers have demonstrated these vulnerabilities across platforms. Vivaldi argues that until these issues are robustly addressed, embedding AI directly into the browser is reckless.

Strengths of Vivaldi’s Approach

A Coherent Privacy Proposition

In a market where data practices are scrutinized, routing user queries through cloud-based LLMs creates new surveillance vectors. By refusing to do so, Vivaldi offers a clean alternative for privacy-conscious users.

Preserving Publisher Value

By refusing to intermediate between readers and sources, Vivaldi shields publishers from the traffic-siphoning effect of AI summaries. This aligns the browser with independent creators and media outlets fighting for a sustainable web economy.

Consistent Brand Identity

Vivaldi has always appealed to power users who want control. Saying no to AI assistants is consistent with a product that lets users customize every pixel and refuses to impose a one-size-fits-all experience.

The Strategic Risks

Feature Parity and User Expectations

AI features are quickly becoming table stakes. Chrome’s ability to summarize a 20-page article in two paragraphs saves real time. The convenience gap between “open a new tab and visit ChatGPT” and “click a button and get the answer right here” is enormous. Vivaldi’s bet that its