Vivaldi CEO Doubles Down on Generative AI Ban
“Web browsing belongs to the people, not the bots.” With that line, Jon von Tetzchner, CEO of Vivaldi, framed a debate that goes to the core of how the internet should function in an era of autonomous systems. In late August 2025 Vivaldi moved from rhetoric to policy, explicitly barring generative AI systems from using its browser to autonomously traverse, scrape, summarize, or republish web content. That stance has reignited questions about consent, copyright, publisher economics, and who — if anyone — should govern automated access to the public web.
Why this matters
For decades, browsers acted as passive conduits: humans requested pages, and pages were delivered. The rise of generative AI has changed that basic model. Modern AI agents can surf the web like human users, follow links, extract text and images, and synthesize answers or summaries at scale. Those capabilities deliver clear benefits: faster research, better accessibility for users with disabilities, and powerful productivity tools. Yet they also enable large-scale automation that sidesteps the social and technical safeguards websites were designed around.
Vivaldi’s objections are clustered around three interlocking concerns. First, consent: sites present interactive consent flows—cookie dialogs, paywalls, personalized prompts—designed for human decision-making. Automated agents can bypass or game these interactions, undermining user choice and the signals site owners rely on. Second, copyright: ingesting and repurposing proprietary text, images, or structured data en masse raises thorny licensing questions. Without negotiated agreements, such ingestion can amount to uncompensated use of creators’ work. Third, publisher economics: if AI-driven summarization and indexing serve users answers without routing them back to original pages, ad impressions and subscriptions can be lost, threatening revenue models that sustain journalism and independent content.
Generative AI and the consent problem
Vivaldi frames the consent problem as central. Generative AI agents, running at scale, can negate the interactive flow that websites use to obtain informed permission from human visitors. That isn’t merely technical quibbling: it has legal and ethical consequences in jurisdictions with strong data-protection rules, and practical consequences for how sites measure engagement and monetize content.
Arguments for autonomous browsing
Proponents of automated agents stress the upside. Generative AI can dramatically speed information discovery, serve as a research assistant that synthesizes disparate sources, and provide accessibility improvements for those who struggle with traditional browsing. Developers warn that forbidding such agents at the browser level could stifle innovation, create a de facto choke point, and fragment the user experience across competing browsers and services.
Policy, regulation, and the limits of private governance
Regulators are already grappling with overlapping legal frameworks: updates to copyright law, data protection rules, and sector-specific regulations. The EU’s deliberations around AI policy and copyright adjustments intersect directly with the question of automated web access. Vivaldi’s move effectively places a governance decision in the hands of a private company, which raises concerns about consistency, democratic oversight, and cross-border enforcement. Should browsers set these norms, or should rules emerge from legislation and multi-stakeholder processes?
Publishers and creators split on strategy
Content owners are not unanimous. Some publishers welcome restrictions that protect ad revenue and bolster bargaining power with AI firms. Others view generative AI as a discovery engine: summaries could lead readers back to original pieces, or foster new attribution and licensing models. The core dispute is over control and compensation—how to enable beneficial AI-driven services while limiting abusive scraping that cannibalizes income.
Technical and legal defenses — and their limits
Traditional defenses like robots.txt or machine-readable signals can convey site owners’ preferences, but compliance is voluntary and easily ignored. Server-side tools—rate limiting, CAPTCHAs, fingerprinting—can slow or deter automated scrapers but at a cost to usability and accessibility. Legal avenues, including copyright enforcement and breaches of terms of service, have been pursued against large-scale data ingestions, but litigation is slow, jurisdictionally fragmented, and still developing relevant precedents.
An adaptive adversary
Determined actors will continue to innovate around defenses. Distributed scraping networks, spoofed user agents, or human-assisted click farms can evade many technical barriers. That dynamic turns the problem into one of continuous adaptation rather than a one-time fix.
A practical test of values
Vivaldi’s ban is noteworthy not because it settles the debate, but because it forces a real-world reckoning between competing values: openness and innovation on one side; consent, control, and economic sustainability on the other. The web’s architecture — built for human interactions — was never optimized for autonomous AI agents, and society must decide whether to retrofit norms, craft clear laws, or accept that private companies will create de facto standards.
Conclusion: who decides the web’s future?
Vivaldi’s stance asks a simple but urgent question: who should decide how the internet is accessed and used — toolmakers, content hosts, AI developers, or the public through democratic institutions? The answer will shape whether generative AI becomes an integrated, consensual part of the web or a contested layer that fragments access and shifts power toward those who control tools and platforms. Without careful multi-stakeholder deliberation that balances innovation with consent and economic fairness, the web risks becoming a battleground where convenience, commerce, and control collide.
