Mozilla’s CTO thinks AI should be built like the internet
Rest of World Itika Sharma Punit ● Covered by 4 sources
Mozilla says AI should be open like the web, not locked behind a few giants. Its CTO thinks businesses are already drifting that way because open models give more control and less lock-in.
Based on reporting by Rest of World, Itika Sharma Punit — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Mozilla has spent decades arguing that the internet works best when no single company owns the pipes. Now it wants the same argument to stick for AI. In a report released last month, the nonprofit said the performance gap between the best open-source models and proprietary systems such as Claude and ChatGPT has shrunk to 3%.
That number matters because the open side is no longer a side show. Mozilla pointed to Alibaba’s Qwen model, which was downloaded more in February 2026 than the next eight models combined. Meta also added fuel to the argument on August 10, when it released a new open-weight model and said more were on the way.
Raffi Krikorian, Mozilla’s CTO, says the shift is already visible in business workflows. Consumer workers may still lean on ChatGPT or Claude, he said, but IT and HR teams are moving toward open models for price, performance, and control. For regulated industries, the attraction is even simpler: keep the data inside the firewall, run it yourself, decide what moves where.
He also thinks the policy debate is stuck on the wrong category. In his view, AI should be treated as infrastructure, not a product that can simply be rented or switched off. That framing matters because the companies shaping the market today are the ones that can most easily push closed systems, while open models give customers more room to self-host, fine-tune, and adapt tools to local needs.
Krikorian is not pretending open source is fully solved. The strict version still requires more transparency around training data, evaluation, and pre- and post-training work. But his bigger point is blunt: if only a few companies define AI, the rest of the world gets whatever those firms decide to sell. If open models keep gaining traction, the rest of the industry has to answer to users, not just shareholders.
My take — AI-written commentary, not fact-checked reporting
The most interesting part of this fight is that it stops being philosophical the moment businesses have to pay the bill. Open AI is no longer about hobbyists and purity tests; it is about leverage. Closed systems may still win the shiny demo contest, but the boring control panel is where the real market usually goes to live.
Read more about this at: Rest of World