The Grassroots Social Network

Connecting through people you trust — with no company in the middle

Child safety is not a policy laid over the network — it is a property of how the network is built.

A social network with no operator

Today's social networks are run by a company that owns the servers, sets the rules, decides what you see, and holds everything you post. The grassroots social network has no such company. It runs directly between the phones of the people using it: you connect to the people you know, follow others, and form groups, and everything you share travels across those connections without passing through anyone's servers.

The difference this makes runs through everything below. Who joins, what you own, how children connect safely — on a conventional platform these are decisions a company makes and can change. Here they follow from how the network is built, and no operator sits above them.

Who is in the network

On most platforms anyone can create an account, and the company is left trying to work out afterwards which accounts are real people, which are fakes, and which are bots. The grassroots social network starts from the opposite place. There are no accounts issued by a central authority; people join the way they meet in life — introduced by someone they both know, who vouches for the newcomer. The network is therefore built out of real, traceable relationships rather than anonymous sign-ups.

The same principle governs AI. An AI agent can take part only when two people vouch for it and are accountable for it: whoever created it, answerable for how it behaves, and whoever operates it, answerable for how it is used — and both of those declarations are recorded and cannot later be denied. At a moment when conventional platforms are quietly filling with AI accounts that pass as people, this is a deliberate and structural difference: on the grassroots social network, no one — human or AI — is present without a real person standing behind them.

There is also no operator to hand control to AI: an AI would have to take over participants' phones one by one. The working paper Corralling AI Risks with Grassroots Social Contracts sets this out. Two other AI risks, harm to people who use AI and harm done by people using AI, are outside the paper's scope.

The things you post stay yours

When you post on a conventional platform, you typically grant the company a broad licence to your words and images as a condition of using the service, and your content lives on its servers under its control. The grassroots social network works the other way around. Every item you post is signed by you and remains yours: publishing it on the network grants no company a licence to it, because there is no company. Your copyright stays with you, by default and by design.

Because each item carries its author's signature, and because everyone who forwards it adds their own, every piece of content carries its own history — a record of where it came from and who passed it on. One consequence matters especially now: the source of a faked or manipulated item is traceable. This does not stop someone from creating a deepfake — nothing can — but it means such content cannot circulate anonymously; its origin, and each person who chose to forward it, remain identifiable and accountable. It also cannot tell whether an item is synthetic, or whether the holder of a key is a person; it shows only who signed it and who is accountable.

As elsewhere on the network, the architecture does not try to police what people say; it makes what is said carry responsibility with it.

Connecting children safely

The same mechanism that builds the network — connecting only through someone who vouches for you — makes it possible to give children a social world that grows the way it does offline.

Young children don't meet strangers on their own; they meet through their parents, at gatherings and playdates arranged between parents who already know one another. The grassroots social network reproduces exactly this. Two children can become connected only when all of the following hold at once: both children want to connect, a parent of each approves that particular introduction, and those two parents are themselves already connected. A child's circle therefore widens outward through their own parents' trusted relationships, just as it does in the physical world — and an unknown adult, or an AI agent, has no way to appear unbidden in a child's social space.

What makes this different from a conventional platform's safety settings is where the safety lives. On an ordinary platform, protections are rules a company applies and can loosen, enforced imperfectly after the fact. Here the befriending rule is part of the software itself: the design cannot form a childhood connection that breaks it, and the working paper paper Child-Safe Social Networking proves this. The proof assumes everyone runs the correct code, which the secure runtime now in development is meant to ensure. Child safety is not a policy laid over the network — it is a property of how the network is built, guaranteed by the same architecture as everything else.

It is worth being exact about what that guarantee covers. What the architecture guarantees is the structure of connection: who can befriend whom, and under whose approval. It does not read children's messages or monitor what they say — there is no operator to do so, by design. And it does not, by itself, verify that someone who approves an introduction as a parent truly is that child's parent; that claim, like responsibility for a child in the physical world, rests on the people involved and, ultimately, on legal accountability rather than on a technical check. For the same reason, a child using their own phone could pass as an adult.

The network enforces structurally what it can, and makes the rest answerable — it does not promise more than it can keep.

Where this work stands

The grassroots social network exists as a tested prototype, released as open source. Its child-safety rule is set out in the working paper Child-Safe Social Networking: From Social Contract to App, which proves that the parental powers hold in every run of the system. The rule has been demonstrated as a phone app with a village of seven people, and in a simulated year of ten families, with 17 parents and 20 children. Like the rest of Grassroots it is at an early, pre-commercial stage: a working demonstration that the design holds on real devices, not yet a product to sign up for. The scientific work behind it is set out in the paper Digital Speech Acts Retain Control of Copyright with People, Not Platforms, and in the draft working papers Child-Safe Social Networking and Corralling AI Risks with Grassroots Social Contracts.

Like the rest of Grassroots it is at an early, pre-commercial stage: a working demonstration that the design holds on real devices, not yet a product to sign up for.

The social network is developed by Ehud Shapiro, with the content-ownership and provenance work with James Golike. The initiative is also supported by Professor Lee Edwards (LSE Media and Communications), an expert in media literacy, public media and public relations in the digital environment.  It builds directly on the Grassroots infrastructure, and runs — like every Grassroots platform — as a mini-app on the grassroots super-app, on an ordinary smartphone.