Public Lecture Series: Transformation and Pressure in Media and Cultural Governance
Dr Giota Alevizou
In October 2024, a US federal court upheld a ruling against the Internet Archive (IA), the organisation that has preserved one trillion webpages and preserves the open web. The case concerned its digital book lending programme. Over 500,000 titles were removed. Within weeks, the Archive was hit by cyberattacks, exposing the data of 31 million users. In 2025, DOGE cut its federal funding. In 2026, at least 241 news organisations across nine countries, including the New York Times, The Guardian, and USA Today, began blocking IA’s Wayback Machine’s crawlers. These organisations fear the Archive could serve as a ‘backdoor’ for AI companies to scrape their proprietary content without compensation.
Wikipedia is facing a parallel pressure. Elon Musk has repeatedly attacked it as a liberal establishment institution and threatened its charitable status. Grok (xAI’s answer to collective knowledge) is being positioned as an alternative. Meanwhile, the GenAI platforms that could not exist without Wikipedia’s training data are quietly diverting the traffic that sustains its volunteer community.
This is not a story about two institutions. It is a story about who governs knowledge, and what happens when the answer changes.
I want to argue that what we are witnessing reveals a structural shift: the moment when the open web, which is built on commons logics and public access, becomes the raw material for a new enclosure. And understanding it requires going much further back than 2024.
Shifting governance regimes
Elsewhere (Alevizou, 2006), I argue that encyclopaedias are not simply reference books; they are knowledge media and, as such, reveal different governance regimes, particularly as they shifted from print to digital, from commercial to open source, from expert credentials to peer production. What made Wikipedia’s procedural epistemology democratic, for example, is just its free from advertising, open access model. It is that the disagreements are visible. Talk pages. Edit histories. Citation trails. Community forums. You can see the friction.
That visibility is, as Matt Vetter argues in a recent article, the condition under which knowledge is co-produced through negotiation and procedural governance rather than institutional decree. It is what makes Wikipedia not just a useful resource but an epistemic infrastructure: one of the most significant experiments in collective knowledge-making in human history.
The algorithmic enclosure
We are now, however, in a period of algorithmic enclosure. GenAI platforms were trained on Wikipedia. The governing norm has shifted from ‘has this survived collective scrutiny?’ to ‘does this sound right?’
Beyond this being a technical shift, it transfers epistemic power from peer to peer communities to algorithms, from accountable processes to black boxes, from the visible friction of democratic knowledge-making to what I’d call the smooth surface of synthetic fluency. Research by Shumailov and colleagues names this model collapse: if Wikipedia becomes saturated with AI-generated content, the quality of AI systems trained on it deteriorates in a negative feedback loop. The commons that trained the machines risk being consumed by them.
This shift carries a structural risk: Contemporary visions of the digital future, I argue in a recent paper, (Alevizou and Gallagher, 2026), frequently sideline democratic politics, often to the point of explicitly proposing post-democratic futures. At the heart of this tendency is the tension between autonomy, as a political and ethical ideal, and automation, as a technological and economic imperative. AI systems automate forms of symbol manipulation and knowledge work hitherto considered the exclusive preserve of humans, and in doing so, they present themselves as open to contingency while locking us into paradigms that privilege speculation over democratic participation. Against both the humanist discourse of rational individualism and the delegation of agency to intelligent machines, we call for forms of futuring grounded in embodied collective intelligence that acknowledge entanglement and alternative imaginaries.
Three simultaneous enclosures
What makes the current moment a crisis rather than a transition is that three forms of enclosure.
GenAI enclosure: LLMs absorb the epistemic labour of thousands of volunteer contributors and knowledge workers, reproduce their authority, and erase their methods. As Robin Mansell has argued, public institutions are positioned as resource providers to AI rather than as co-governors of its development.
Legal and commercial enclosure: Publishers restrict web crawling. The Internet Archive loses in court. The EU AI Act requires training data transparency for general-purpose models (a genuine achievement), but its opt-out mechanism favours large commercial rights-holders over individual contributors and commons-based infrastructures. What happens when the people who built the open web lack the administrative capacity to protect it?
Political enclosure: Musk’s attacks on Wikipedia are structurally significant. Wikipedia’s authority derives from decentralised verification. That is precisely what makes it a target. The effort to delegitimise it is an effort to replace accountable epistemic infrastructure with one that cannot be contested.
Custodianship as the alternative?
We can be reminded of political philosopher Hannah Arendt’s argument that the public realm depends on plurality: on the irreducible fact that it is this person, in this position, making this claim, who can be held to it. Wikipedia is , perhaps, one of the few remaining large-scale spaces where that kind of attributed, contestable, accountable knowledge-making still exists. What AI enclosures threaten goes beyond information quality; they may lead to the vanishing of the ‘who’ behind the ‘what’.
We may be reminded of Tarleton Gillespie’s custodianship argument that content moderation and platform-mediated speech [e.g. on Facebook] receives too little public scrutiny. Social media platforms are powerful custodians who pretend they are not, and the custodianship tactics are motivated by commercial (and political) logics. But the more urgent problem today is the opposite: institutions that were legitimate custodians (e.g. Wikipedia, public broadcasters, the Internet Archive) are being progressively dispossessed of that role while extractive systems absorb their authority without their accountability.
The governance response to this demand goes beyond content moderation or even training data transparency. It requires what I call — drawing on Joan Tronto’s ethics of care — an institutional practice of custodianship: long-term relational obligation to the conditions that make knowledge possible. Cultural memory. Epistemic diversity. The labour of the people who maintain knowledge infrastructures. The ecological costs of the compute power that runs them.
Current frameworks are not adequate to this task. The EU AI Act says nothing about epistemic infrastructure as a public good. It has no provision for reciprocity obligations for AI trained on commons-based knowledge. It does not position public knowledge institutions as co-governors of AI development.
What this requires
AI is not going away. The question is not resistance or capitulation. It is, as Michel Serres would have put it, what kind of relation with the parasite remains liveable — and on whose terms. The answer requires three things that current governance frameworks do not provide:
A positive legal status for epistemic or public media infrastructures (similar to what we have for public service broadcasting) that treats platforms like Wikipedia, the Internet Archive, and public archives as public goods, not resources available for extraction.
Provenance and contestability as governance standards. The right to know not just what AI says but where that knowledge came from; this includes the right to push back. Public knowledge institutions as co-governors of AI development, not feedstock. The communities whose labour built the open web should have a voice in how it is used.
These, I think, are the conditions of epistemic justice. And they are, right now, more fragile than they have ever been.
Dr Giota Alevizou
Lecturer in Digital Humanities and Culture, King’s College London
Director, MA Digital Futures, Author of The Web of Knowledge: Encyclopedias and Authority in the Digital Age (Cambridge: Polity, 2026)
This post on draws on a public lecture delivered at the Media Governance and Industries Research Laboratory, University of Vienna, 28 April 2026, as part of the series ‘Transformation and Pressure in Media and Cultural Governance’, convened by Prof Katharine Sarikakis.
Leave a Reply