Mark Zuckerberg says the future of AI is for everyone. But who owns it?

On Monday, Mark Zuckerberg published a 6,500-word essay, The Future Is for Everyone. The essay came with something even rarer: a new Meta open-weight model. An open-weight AI model is the kind of AI with which you download the entire thing, and it works on your own computer (with or without the internet) and nobody can switch it off except for you. Almost every mainstream AI tool you use, whether it be Gemini, ChatGPT or Claude, is one you rent access to. Meta’s is one that you can download and actually own. I downloaded it before I finished the essay. So can you.

This major move by the world’s largest tech heavyweight comes amid a summer heavy with AI news, full of nations at war. Washington against Beijing; tech founder against tech founder. But beneath those conflicts is a deeper contest between two kinds of power: a government that can order technology offline in an instant, and a company that can simply close your account. Neither is necessarily villainous. But the rest of us are not merely the audience in this fight; we are the prize. And the terms are already being written: intelligence on lease, a future we pay for but never quite own.

Before you sign a lease, you generally wonder three things. Can I be thrown out? When I go, what goes with me? And if the landlord loses interest, who keeps the place standing? Ask the same three questions of any technology that says it belongs to everyone.

Start with eviction. In 2009, Amazon reached into people’s Kindles overnight and deleted a book that everyone who bought it thought they owned. Nineteen Eighty-Four. On 12 June of this year, the US government did the same thing to AI: three days after the American lab Anthropic launched its most capable model, an order from Washington forced it offline for everyone, everywhere. Within days of Anthropic’s release, I tried two Chinese open-weight models, Kimi K3 and GLM 5.2. I could download them. I could run them myself – and nobody could reach into my server and delete them. On several of my tasks, they performed nearly as well. As such, the deciding question on open or closed models is no longer capability; it is who has the power to switch it off.

The second question is one Zuckerberg’s essay never asks: can you take it with you? The essay does make a real privacy commitment. It introduces a fully private mode, sealed the way WhatsApp is, where even Meta can’t reach. The landlord will be locked out by architecture, and not simply by a promise. But read carefully. The agent he describes will hold your health, stories of your children and years of your thoughts – all in a diary kept in someone else’s house. Ask again: can you take it with you?

And the third question: when the landlord loses interest, who keeps the place standing? Zuckerberg says Chinese models are leading on open-source AI because American labs face heavier data rules. That’s not quite the full story. Chinese labs open their models because openness is their business model. American labs, on the other hand, keep theirs closed because openness would undercut them. Meta was the exception – until it paused its open strategy last year. But then something remarkable happened: absolutely nothing. Everyone who had downloaded Meta’s models kept them, and just kept building. Companies can change their minds, but they cannot undownload a model from somebody else’s machine. The files people had saved kept the ecosystem alive, not the company’s goodwill.

We’ve seen this pattern before. Decades ago, one company gave a free browser to billions – and that act almost closed the web. Internet Explorer was “for everyone” too. What saved the open web was not a better company stepping in, but software built to work for the use, rather than whoever made it. Mozilla exists because of this fight. And this current battle over AI is the sequel.

So everyone using AI should have three demands – and none of them is “pick a better company to trust”.

Hold. Washington already gets this – when Washington itself is the tenant. Its own security posture warns against AI that an outside company could easily switch off. Britain, Canada and Germany already spend significant money on sovereign computing. Extend these principles to civilian life, and critical services like hospitals and government agencies should be required to hold legally usable copies of capable open-weight models, along with the tools and computing capacity to run them. All should face rigorous security and safety testing, whether they come from China, the US or anywhere else. Test the capability, don’t ban the method.

Take. Demand of agents what we just won from models: portability. The memory a machine has of you should be exportable to a rival. The diary should not be trapped in the house. This is the fight that will define the decade.

Build. Openness shouldn’t depend on one company’s willingness to keep sharing. We need to pool the nuts-and-bolts of open-source infrastructure – the testing labs, the harnesses that run the models and the standards – so that the open ecosystem has what the web has today: a foundation that no single company can switch off.

Zuckerberg’s title is right: the future should be for everyone. Open models don’t eliminate dependency; they give us what digital infrastructure rarely provides – a real, practical way to walk away.

• Raffi Krikorian is the CTO of Mozilla