Digital public service infrastructures for culture

Is a non-commercial mandate possible?

Speech to Oxford Internet Institute, 18 June 2026

On 18 June I spoke at the Oxford Internet Institute, at a seminar shared with Dr Ignacio Gallego of the Universidad Carlos III de Madrid, held at Oxford's new Schwarzman Centre. The session posed a single, seemingly futile question: whether it is any longer possible to build genuinely non-commercial digital infrastructures for culture at a time when almost all of our cultural life is produced and circulated on corporate platforms whose business models extract personal data in return for visibility.

I was asked to answer that question using dotPublic as a case study, and to set it against the wider debates about digital public infrastructures, digital public spaces, and the future of public service media. I argue that the public service mandate should reach deep into the infrastructure of the internet itself, and I reference the visionary work of the engineer Jens Finkhäuser, whose insight is that the pieces needed to build a public service layer already exist, albeit somewhat scattered across the internet, and therefore do not need be invented so much as intentionally connected.

---

Thank you. And thank you to Ignacio for setting out the territory.

The question in the headline is whether a non-commercial digital mandate for culture is possible online. It is a timely question and an important one, and I want to frame it seriously before I attempt to respond to it.

The most straightforward response is that a non-commercial mandate is possible in exactly the same sense that a non-commercial road is possible. A road can be built and maintained by public authorities, regulated by public law, used freely by anyone and everyone, and judged simply by whether it gets people safely from one place to another.

A public road does not prohibit commercial traffic. It simply serves a different purpose.

Lorries can use it. So can buses, cyclists, ambulances, and ice cream vans. What makes the road public is not the absence of commerce but the presence of a different organising purpose. The road exists to serve the people who use it, and the companies, organisations and institutions that depend on it, on terms set to serve the public good.

The internet, in its current form, is not such a road.

It is a road system that is extraordinarily good at some things. Capturing attention. Collecting behavioural information. Targeting individuals. Optimising for engagement. These are remarkable achievements, and I do not want to be heard to be dismissing any of them.

But they are not the objectives around which our cultural and civic institutions were founded.

Libraries were not created to maximise engagement. Museums were not created to optimise behavioural outcomes. Public service broadcasters were not founded to harvest data. Archives were not built to maximise dwell time. They were built to inform, remember, safeguard and preserve.

The question I want to put before you this afternoon is whether public institutions can continue to fulfil these public purposes on digital infrastructure designed for entirely different ones.

That is the question that has occupied me for much of the last thirty years, working inside the Guardian, the BBC, and the New York Public Library. What I have observed in those institutions, as well as in the public-serving organisations I have worked alongside, is not a failure of purpose. The librarians have not changed. The journalists have not changed. The curators have not changed. More importantly, their public mission has not changed.

What has changed is the infrastructure through which that mission is now being delivered. And infrastructure, as it turns out, has defaults. Not cultural ones. Structural ones.

All infrastructure embodies assumptions about what success looks like, and in time those assumptions have consequences.

I want to introduce you, if you do not already know him, to the Irish writer, Flann O’Brien.

In his comic novel, ‘The Third Policeman’, written in the 1940s, O’Brien sets out what he calls the molecular theory of bicycles. The idea is that pressurised contact between two things causes them gradually to exchange their atoms. Just as bashing a hammer onto concrete leaves both fine particles of iron on the concrete, and tiny fragments of concrete on the hammer.

In his story, the man in question has spent so many years bumping along on the same bicycle that he has, over time, become forty-eight per cent bicycle, while the bicycle has become forty-eight per cent man. They have begun, gradually and in plain sight, to become indistinguishable.

It is an absurd idea. It is also one of the best descriptions I know of what has happened to public institutions online. For twenty years public institutions have been continually pressurised by commercial platforms and proprietary infrastructure.

That pressure has been applied through procurement frameworks, consultancy advice, industry norms, software defaults, and the received wisdom that commercial tools are easier, and seemingly cheaper, to adopt than building public alternatives.

Under that constant pressure their positions have begun to exchange. Public institutions use the same optimisation tools, the same behavioural frameworks, and the same definitions of success, as commercial operators. They have become, by some measure, forty-eight per cent platform.

At the same time, the platforms too have become increasingly adept at presenting themselves as public serving. They speak in the language of community. The language of personal service. The language of public value. They adopt the visual and rhetorical forms of the institutions that spent centuries earning legitimacy, and the public’s trust.

The result is that ordinary people can no longer easily distinguish between the two. And as distinctions disappear, so trust begins to erode. Not because the institutions changed their intentions. Because they changed their infrastructure.

For a long time I thought the central failure of the internet was misinformation. I no longer think that. The deeper failure here is one of provenance and permanence.

For roughly five centuries, since printing made the book a mass object, publishers developed mechanisms that helped people answer four remarkably simple but essential questions.

Who created this? Under what authority? Has it been changed? Will it still be available tomorrow?

Those questions were embedded within published documents. On title pages. In publishers’ imprints. In editions and citations. In archives, libraries, catalogues, and footnotes. In records of custody.

The internet solved distribution brilliantly. It solved provenance not so much. Information has become easier to publish than at any time in human history, but the mechanisms that told us where it came from, who stood behind it, how it had changed, and whether it could be trusted, often disappeared in the process.

We now find ourselves in a world where more information is available than ever before, and yet many of us are less certain than ever about where any of it originated.

This problem becomes considerably more significant as artificial intelligence becomes a primary interface to knowledge. Initially, with the creation of the web, we were building systems primarily for humans to read. Increasingly we are building systems for agents. Agents cannot evaluate using instinct. They cannot rely upon reputation in the way humans do. They depend instead on structured data, machine-readable signals, verifiable chains of authority, and ways to distinguish between information that merely exists, and information that can be traced to an accountable source – an author, a publisher, an institution.

That observation is one of several roads that led to dotPublic. Privacy is another. So are accessibility, transparency, and the question of what persists. Each leads to the same place.

And so dotPublic begins with a simple proposition. Public institutions require public infrastructure. Not public ownership of everything. Not public control of everything. Public infrastructure. Designed around curation rather than extraction. Provenance rather than popularity. Transparency rather than opacity.

The work is anchored at Trinity College Dublin, where the dotPublic research programme runs in the ADAPT Centre under the principal investigation of Professor Dave Lewis, and where I hold an adjunct appointment. It is built around three things that work together.

The first is a framework of nine civic standards. In no particular order they are accessibility, privacy, security, transparency, accountability, interoperability, provenance, AI and automation reporting, and responsibility to the future. Having said that, I mention accessibility first because it is so often considered last.

Otherwise, the other standards are not ranked. They are foundations, and each one strengthens the others. They are not a manifesto. They are obligations, expressed in language that public institutions already recognise, and identify with, and are structured so that compliance can be evaluated.

The second is a new top-level domain, governed by those standards. I cannot name the string this afternoon, but I can tell you what it does. It creates a bounded civic space. Inside that space, the standards are not aspirations. They are conditions of participation. Outside the civic domain, the rest of the internet continues exactly as it is.

The third is decentralised infrastructure for continuity. Public records have to outlast the platforms that hold them, the suppliers that built them, and in some cases the institutions that originally created them. Earlier this year, working with Trinity’s ADAPT Centre, we reconstructed more than four thousand records from twelve years of the Centre’s research in a weekend, at a cost of around four thousand euros.

The exercise was not technically remarkable. What it demonstrated is that the obstacle to digital continuity is no longer cost or capacity. It is the absence of an infrastructure designed for the purpose. This brings me to what I find most remarkable about the whole proposal. The longer we have worked on it, the more apparent it has become that almost nothing dotPublic asks for is new.

The commitments to public service already exist. Accessibility, privacy, accountability, the diligent stewardship of the public record – these are not new obligations. They are the founding principles of the institutions themselves, written into their charters and their practice long ago, yet quietly let slip as public life moved onto digital infrastructure built for other ends. And the technology already exists. Through DNS, the internet already knows how to name institutions reliably.

Through DNSSEC, it already knows how to pass authority from one body to another without losing track of who is in charge at each step.

Through attestation frameworks, it already lets organisations make signed statements about themselves that anyone can check.

Through persistent identifiers, like DOIs, it can already give publications permanent addresses that outlast websites and redesigns.

Through federated preservation networks, it can preserve records across many places at once, so the failure of any one institution does not erase its work.

Through transparency logs, it can keep public ledgers of important changes, so nothing can be quietly rewritten.

Many of these building blocks were developed in their own corners of the internet over decades, by engineers and archivists and standards bodies who may rarely have been in the same room, and whose tools have seldom been brought into conversation with one another.

What no one had quite seen apparently, was that they could be made to work as one. That was the insight of the German engineer Jens Finkhäuser: not to build anything new, but to connect what already existed into a single delegated chain of trust.

The challenge, in other words, may not be reinvention. It may simply be reassembly.

That is a very different proposition from building an entirely alternative internet. It is closer to building public roads on land that already has well-trodden paths running across it. The objective is not to replace what is there, but to ensure that some part of the existing environment becomes reorganised around public purposes, and that the institutions which serve those purposes have somewhere to operate from.

So, to return to the question: Is a non-commercial mandate for culture possible online?

The word 'mandate' is doing a lot of work in that question, and could mean a number of different things. A democratic mandate, conferred by elections. A legal mandate, set by statute. A constitutional mandate, written into the founding charter of an institution.

Each of these meanings ends up in the same place. An obligation, willingly undertaken by the entities entrusted with public information, goods and services, to operate in a manner that is in the best interests of the public, at all times and without reservation. Read that way, the question becomes a practical one. Is it possible to build the infrastructure such an obligation requires?

I think the answer is yes. A mandate, or obligation, such as that, without the supporting infrastructure is a wish. And a wish, expressed from deep inside an environment optimised for other goals, will gradually become indistinguishable from the environment around it. Forty-eight per cent platform. Then forty-nine. Then fifty.

And from there on the question stops being how much platform we have become, and starts being how much of ourselves remains.

The internet could have become the greatest public information resource in human achievement. Instead it became a shopping mall built largely around the capture and monetisation of attention and behavioural information. And disappearing somewhere inside that system are our libraries, archives, and the cultural institutions we built to serve very different purposes.

The institutions we are discussing today were created because previous generations believed some things were too important to leave entirely to markets. Knowledge. Memory. Culture. The internet did not remove those obligations. If anything, it made them more important.

The question is not whether a non-commercial mandate for culture is possible online. The question is whether we are prepared to build it.

Thank you.

---