· Link

Who does the agentic web belong to?

W3C’s new proposal strengthens the case for a community-governed way to agree how agents use the web’s content and services.

On 21 October, the World Wide Web Consortium published a Strategy Team discussion paper proposing duties to users, meaningful consent, interoperability and machine-readable terms for content training and reuse. These are proposals for discussion, not an adopted standard or an agreed position of W3C’s membership.

I find this encouraging. The value-exchange problem I described in The agentic layer needs a practical answer: who sets the terms when an agent uses someone else’s work, and what does the source receive in return?

That is why we are working on odexa, an open protocol for content-use permissions, agreements and evidence of participating exchanges. I am its Foundation Chair. This proposal strengthens the case for that kind of solution, although W3C has not endorsed odexa.

Making it work is a huge undertaking. My commitment is to solving the problem. If the web community gets there through odexa, another protocol or work that brings several approaches together, that is progress. What matters is that it works and remains in the community’s hands.

Companies such as Google, OpenAI and Anthropic should be able to participate without any one of them controlling the rules. Community governance and commercial participation can coexist. Publishers and developers should be able to work across providers without having to accept a different bargain at every boundary.

Interoperability matters here. W3C points towards existing work on rights and permissions, including ODRL and TDMRep. odexa needs to fit into a wider web of standards, rather than become another isolated system.

That bargain also carries responsibilities for originators: originality, accuracy and quality. We should be clear about what we created, credit what we draw from, substantiate our claims and correct mistakes. Permission to use something is not evidence that it is worth using.

A web flooded with copied claims and recycled errors is a poor foundation for systems that learn from it or use it to answer questions. Sustaining good source material is part of making the agentic web work. Originators must take responsibility for what they publish, and model providers for how they select, check and use it.

The hardest part will still be getting agents and their operators to honour agreements. W3C has the standing to bring relevant organisations together. Turning that influence into implementation commitments would make this significant. A published standard alone would not establish compliance.

It is also a test of W3C’s role in the next web: can it help establish shared rules quickly enough to shape how agents operate?

I call this emerging intelligent and agentic web Web iv. My forecast is that, by 2030, delegating everyday tasks across websites and services will become normal. The Web iv clock records the evidence for and against that forecast; it does not yet calculate an arrival date. This proposal adds a standards milestone, not evidence of widespread adoption. My forecast remains 2030.

The next checkpoint is the paper’s scheduled discussion at W3C’s Web and AI Interest Group session on 29 October.

W3C Strategy Team

Event: 21 Oct 2026 Status: announcement Evidence source