Tim Berners-Lee’s plan to check Big AI: Solid data pods

TL;DR:

Berners-Lee argues for user-owned data “pods” via Solid.
His book and recent interviews set the case in 2025.
Pods can feed AI agents without handing data to platforms.
Governance and standards decide if this works at scale.
Early pilots show promise, but network effects are hard.

Tim Berners-Lee, who created the web in 1989, is again pushing for a reset. In 2025 interviews and a new book, he warns that AI built on platform hoards of data will erode privacy and trust. He proposes Solid, a standards-based way to store personal data in private “pods,” then grant time-bound access to apps and AI agents.

The problem he is trying to fix

Today, a few firms control the data that trains and runs AI. That centralization shapes what models learn and how services behave. Berners-Lee says this breaks the web’s original user control. His September 2025 media tour ties the web’s drift to surveillance, ad targeting, and engagement-heavy feeds. He argues AI now raises the stakes.

The Solid idea, in plain terms

Solid separates data from apps. Your profile, health logs, purchases, documents, and social graph live in your pod. Apps ask for specific slices, for a set time, with your consent. You can move the pod or revoke access. Inrupt, his company, ships the enterprise tools to make pods work across institutions.

How this helps with AI

Training and inference on your terms. A personal AI agent could read your calendars and messages from your pod, not from a company vault. It acts for you, then forgets when access ends.
Data minimization by default. Models and services only see what they need. Fewer copies mean fewer leaks.
Portability. You can switch providers without losing history, which weakens lock-in.

What is new in 2025

Berners-Lee’s book “This Is For Everyone,” published in September 2025, lays out a manifesto to reclaim the web’s open spirit and to back Solid as the technical path. TV interviews the week of September 17, 2025, broaden the pitch to AI agents that “work for you, not for a platform.”

Researchers are also debating standards-driven approaches to AI. A recent paper argues private, standards-based data layers can align incentives better than platform fixes, echoing Solid’s goals.

Where this could work first

Health and social care. Pods can unify records across clinics, then feed care AIs with consent.
Education. A student’s portfolio follows them, while tutoring AIs read only selected items.
Civic services. Cities can build apps that read resident pods instead of building new silos.

These domains already face strict privacy rules, which fits Solid’s consent model.

Real limits to watch

Adoption friction. Big platforms live on centralized data. They may resist deep pod support.
Usability. Consent UX must be simple, or people will click “allow all.”
Security. Pods concentrate valuable data. Hosting and key management must be strong.
Standards governance. Solid wins only if many vendors interoperate in the open.

What success looks like

App stores require granular pod permissions.
Major clouds offer managed pod hosting.
AI agents default to pod access, not platform caches.
Regulators recognize pods as a valid path to data access rights.

Progress on any two of these would shift power away from AI gatekeepers.

Quick checklist: how to pilot Solid in 2025

Start with one narrow use case, like calendar plus notes.
Host pods with an enterprise provider, then test migrations.
Build consent flows that expire by default.
Log every data read for audit.
Measure switch costs between two Solid-enabled apps.

Why it matters

AI value comes from context. If users hold context in pods, they can bargain. That pushes vendors to compete on service quality, not lock-in. Berners-Lee’s plan is not a silver bullet, but it is a concrete lever that technologists and policymakers can pull now.

Sources: