Reflections from MozFest 2026 in Barcelona

First, thank you

I spent last week in Barcelona at the Mozilla Festival, representing the GLIA Foundation’s GliaNet initiative, along with my colleagues Estefanie Govea and Nell Clasby. And I came home to the San Francisco Bay Area holding two feelings that do not quite fit together. I want to briefly set them side by side here.

The first feeling is hope. For three days at Fabra i Coats, a former textile factory in Sant Andreu, I was surrounded by people who have decided not to wait for permission to build the world they want to live in. Activists and technologists. Philanthropists and librarians. Researchers, organizers, artists, public servants, and a great many civic-minded citizens who hold no title at all. Some of them have been at this since the early days of the open Web. Others arrived only recently, pulled in by what AI is now doing to their work, their communities, and their children.

MozFest's theme this year was Wilding: a call to rewild the Internet, to restore balance, to cultivate infrastructures rooted in care, mutuality, and collective action. That is not a bad description of GliaNet’s core mission with our fiduciary AI governance framework. It was a gift to spend the week among people who already understood why such things matter.

To everyone I met there, and to the many more I did not get to: you are the reason the first feeling is hope.

Pick your poison – or don’t

The second feeling is harder to name. Call it a growing sense of clarity, or perhaps alarm. Or both.

While we were meeting in those welcome pockets of human innovation in Barcelona, the larger shape of AI governance has been continuing to harden around us. In recent months we have watched a new kind of oligarchy taking form between the United States and China. Two countries, two models, and a world being told it must choose between them.

On one side, the China model, the state defines what AI systems are for, sets the firm limits of what they can say and do, and leaves private companies to build within those limits. On the other side, the American model, the largest tech companies define what AI systems are for, while the current Federal government works to clear their path from any meaningful oversight (including from forward-thinking States). The two countries obviously differ in outlook, in culture, in history, and in a thousand particulars that matter. But strip away the particulars, and this is the choice being presented to the rest of us by the current leadership: a handful of tech executives determining what the machine will sell, or nameless bureaucrats deciding what the machine may say. One openly rejects any broad-based global efforts to regulate AI; the other embraces the idea so long as they will be wielding one of the pens. In other words, pick your poison.

This doesn’t seem to me like hyperbole. Consider that Tom Friedman, back from nine days in China, devoted his Sunday column to welcoming exactly this outcome. He describes a “two-superpower marriage” between Washington and Beijing, and welcomes the prospect that the two will jointly develop and manage AI. The trigger, he says, may have to be a crisis large enough to frighten both, but not quite large enough to end our world. A “perfect A.I. storm,” in his view, just short of catastrophe, can help produce a path to agreement.

Set aside how it sounds to be hoping for such disasters. It is always the case that crisis-driven rulemaking is incumbent-driven rulemaking. Written by the already powerful, for the already powerful. So while Friedman may well be right about the forecast, the question he doesn’t ask is: who speaks for everyone else?

I think I can fairly say this is not the kind of choice many of us are eager to accept. Both systems offer obligations imposed on AI companies. The problem is not an absence of duties. It is that every one of those duties runs either from the companies to the state, or from the companies back to themselves. Neither of those options puts the actual human being at the center of the equation. Neither builds in a general duty of care, or of fidelity, to the people the technology is supposed to serve. Both would treat the rest of humanity as a captive audience, or a closed market, or an existential risk to be managed — rather than as the actual set of principals whose interests should matter most.

Indeed, a generation or two ago, first the open Internet and then the open Web were designed and built by people who refused a similar choice of entrenched governmental control, or shareholder fiat. Today, we should refuse this one as well.

We demand Plan C

Now is the time to stand up and tell the world, plainly: this is not what we want.

Let’s call the ready alternative Plan C. A third path, in which the technology represents each and every one of us, and answers to us, rather than to a ministry, or a boardroom. Where the AI systems that increasingly mediate our lives owe us the same duties of care, fidelity, and loyalty that we already demand of doctors, lawyers, and trustees. Where the people are not the product, and not the subject, but the principal.

This pro-human option is not a utopian ask. It is a governance choice. And pieces of it already exist in the pockets I saw in Barcelona: open standards and protocols, community-owned infrastructure, personal agents built to serve their users, cities insisting on technological sovereignty.

Nor does Plan C need to gain permission from Washington, or Beijing. The so-called “middle power” countries, those democracies in Europe, Canada, and across the Global South, have the most to lose from a two-bloc AI world. As well as the most to gain from a unified alternative. They have the policymaking weight, the procurement power, and the moral standing to set a different standard – if they choose to use it. Perhaps that standard could be a mix of diverse tech supply, and common governance standards, paired with meaningful virtual defenses for their own people.

It turns out we now have a date on the calendar. Next June, governments of the world and industry representatives will gather in Switzerland for the Geneva AI Summit. Perhaps this will be the last broad multilateral conference table on AI policy, before the bilateral one is set. Nonetheless, the road from here through Geneva, and beyond, is an opportunity for all of us in these singular pockets of resistance to pull together, to plan, and then to arrive. Not as a scattering of worthy projects, but as a single, vocal demand for something better. The pro-human AI stack.

From pockets to a movement

Here is the challenging part. What I saw in Barcelona was hugely inspiring — and it was also relatively modest. Two hundred sessions, a few thousand people, a great many good ideas that will go home to separate inboxes. Pockets are how movements begin. They are not how movements win.

The current geopolitical AI duopoly seemingly has all that it needs: political power, gobs of money, immense scale, and the default on its side. To respond in kind, we will need the proverbial big tent: a coalition broad enough to hold the open-source builders and the digital-rights lawyers, the privacy advocates and the librarians, the labor organizers and the foundations, the city governments and the middle powers. All pulling toward the same essential standards. That means collaboration across projects that have, until now, worked mostly in parallel. It means coordination between governments often searching for their own bespoke answers. And it will require resources sufficient to turn this from a small, admirable corner of the ecosystem into an actual social movement.

GliaNet will do its part. We believe in rules of the road for AI systems that reduce the growing power asymmetries, rather than entrench them. And, AI agents that bolster our human capabilities, rather than siphoning them off. We plan to keep building our fiduciary infrastructure: the Net Fiduciary™ open standard, the personal and community augmentation agents, the technical specs, and the policy frameworks that can help give Plan C a concrete shape. And we plan to show up, in Geneva and in the road leading there, ready to work with anyone who shares the goal.

But no single organization can carry this. If you were in Barcelona, or wish you had been, and you are working in one of these pockets of pro-human innovation, we would like to hear from you. The large tent is being raised. There is plenty of room inside.

Castells

On one evening of the festival, a Catalan colla built castells in the courtyard. These are human towers, several stories high, constructed by raising one person at a time on the shoulders of the people below. No one at the top gets there alone. No one at the base is there for the view. It works only because every person in the structure has decided to trust the others, and to bear the weight.

I can’t think of a better picture of what we are trying to build.

We want Plan C. Thank you, Barcelona, for the reminder of how many of us already do.

Sources

• Thomas L. Friedman, “The Big Change Coming in U.S.-China Relations in the Age of A.I.”, The New York Times (Nov. 2, 2026).

Want to take this further?

If you were at MozFest 2026, or are working in one of these pockets of pro-human innovation, we would like to hear from you.

Contact Us