What the Commons Built (And What's Taking It Apart)

In 1976, Bill Gates wrote an open letter to hobbyists accusing them of stealing. What they were actually doing was sharing software they had written for each other (modifications, tools, documentation), the way people had shared knowledge since the first person showed another how to do something useful. Gates reframed mutual aid as intellectual property theft. It was not a philosophical claim. It was a property claim, backed by lawyers, Congress, and eventually the World Trade Organization.

What happened next is worth understanding clearly, because it is happening again, faster and at much larger scale.

How the Commons Began

Software was born as a commons. Not by design, not through policy, not because someone decided it ought to be. It grew from the implicit practice of early computing communities: hardware was expensive, expertise was scarce, sharing made the whole system work. The people who wrote the first programs shared them the way scientists share data or cooks share recipes, because hoarding knowledge makes everyone worse off, including you.

This commons had no name, no legal structure, no formal governance. It was just what people did. By the time Gates noticed it, it had already become foundational to a significant portion of the world’s computing infrastructure.

The response, Richard Stallman’s free software movement, was a precise legal countermove. Stallman used copyright’s own mechanisms, twisted into a “hack,” to protect the right to share. The General Public Licence (GPL) guaranteed that any code released under it had to remain shareable. You could build on it, modify it, distribute it, but whatever you made from it had to be made available on the same terms. Redistribution was the lever; the licence rode downstream on every copy.

By the late 1990s, this had become a global ecology of commons-based production: Linux, Apache, Wikipedia, OpenStreetMap, the Budapest Open Access Declaration, Creative Commons. Projects that demonstrated, empirically, that complex knowledge goods could be produced and maintained at extraordinary scale without either market pricing or corporate hierarchy. The commons worked, and it worked better than most people predicted.

Three Ways the Commons Got Absorbed

Here is where the history gets uncomfortable.

The digital commons did not need to be destroyed. It needed to be absorbed, partially, strategically, in ways that preserved its usefulness while neutralising its potential to challenge the logic of extraction. Over the past two decades, three methods have done most of this work.

The fix. When markets fail to provide something, or when public services have been cut, the commons gets called in to fill the gap. Wikipedia substitutes for the reference libraries that austerity eliminated. Open mapping fills the gap left by under-resourced public agencies. Crowdsourcing replaces paid research labour. The commons becomes a shock absorber for market and state failure: reliable, available, free-at-point-of-use. It stabilises the system rather than challenging it. When Margaret Thatcher’s successor David Cameron promoted “big society” and voluntary organisations to substitute for the public services austerity was cutting, he was articulating what capital had already discovered: the commons is convenient when you need something for free.

Pre-competitive cooperation. The Linux Foundation is funded almost entirely by industry: Platinum membership costs $500,000 per company per year, with Google, Microsoft, Intel, and Meta among its members; total annual revenue exceeds $300 million. The same companies funding it in the pre-competitive phase lobby fiercely against open standards wherever those standards threaten their competitive position. They support the foundations everyone needs, the shared plumbing, while fighting against open approaches that might disrupt their market advantage. The commons becomes an R&D subsidy, paid for in the open, with the resulting competitive insights used in proprietary products. As one analysis of this dynamic puts it: the perspective and practice of pre-competitive cooperation draws the boundaries between public, private, and common to serve capital.

Enclosure. The most pervasive form. In the 1980s, the GPL’s requirement that redistributed software must remain free was intended to prevent enclosure. Cloud computing changed everything. If you run code as a service, software that lives in a data centre and is accessed via a browser, you never technically “distribute” it. The licence’s trigger never fires. Google’s systematic closing of Android is the canonical case: de facto proprietary, de jure open. Free software has been combined with closed services so thoroughly that the letter of the licence can be satisfied while everything that made the licence meaningful has been extracted.

Social media went further: it pulled cultural commoning itself onto private platforms. The commons of human connection, dialogue, and creative collaboration was enclosed and turned into a surveillance and advertising business. As Felix Stalder at Zurich University of the Arts notes, social media platforms rely on commoning as a bottom-up popular practice, and then use that practice to extract data at scale, the containment not just of data but of human activity itself. What remains operates, as he puts it, “in corrupted, crippled forms.”

What AI Is Doing That’s Different

The previous enclosures reduced the potential of the digital commons while leaving the everyday practice largely intact. AI is now threatening the everyday practice itself.

The argument runs in three parts.

First, the expropriation. Large-scale AI systems require enormous amounts of training data. The digital commons, with its built-in mechanisms for access and sharing, became an ideal training resource: freely accessible, human-generated, high-quality, already organised. Decades of Wikipedia editing, OpenStreetMap contributions, open scientific publishing, and free software documentation have been consumed as training inputs. The commons licenses allowed this: they were designed for sharing, not against AI consumption, which didn’t exist when they were written. The result is that billions of hours of human collaborative labour have been transferred, uncompensated, into the training infrastructure of commercial AI systems.

Second, the pollution. Generative AI is flooding the commons back with what has been called “AI slop”: plausible-looking but factually unreliable content at scale. Wikipedia’s quality control has depended on human editors applying human judgement to distinguish reliable from unreliable contributions. AI-generated content has the same surface form as human contributions: the same stylistic confidence, the same apparent specificity. A representative from the Wikimedia Foundation, speaking at a European Commission workshop on generative AI and the digital commons in March 2025, put it plainly: “In the past, low-quality content looked like low-quality content. Now it looks like high-quality content.” The heuristics collapse. The editorial labour required to maintain quality explodes. Smaller language editions of Wikipedia are already struggling.

Third, and most fundamentally, the isolation. Commoning is a social practice. It requires people doing things together: editing, correcting, discussing, contributing. Each person locked in a personalised chat window, receiving individualised answers to individual queries, is being isolated from the community whose work they are consuming. AI chat obscures its sources, provides no path to collaborative contribution, and suppresses the traffic and engagement that makes it worth contributing to the commons in the first place. Click-through rates from AI-generated answers are around 96% lower than from traditional search engines. The social infrastructure that sustained the commons is being monetised and then starved.

What Survived, and Why

Three commons-based projects have most clearly escaped this dynamic: Debian, Wikipedia, and OpenStreetMap. Each operates outside the market and the state, in complex institutional arrangements where informal communities, not-for-profit organisations, and commercial firms can work together under community-defined guidelines that put the community’s interest first.

What they share is timing. Debian built its community culture before free software became a tech industry strategy. Wikipedia and OpenStreetMap were initiated before commercial social media established data extraction as its core business model. The communities that might have been captured were built before capital arrived with attractive offers.

This is not a coincidence. It is a lesson about governance. Once the commercial logic is embedded at the foundation, once the entity is structured to serve investors rather than members, the community that might have resisted it never fully forms.

It also means something uncomfortable: no comparable digital commoning project has emerged since. The conditions that made these projects possible, the window before commercial colonisation, the availability of builders who were not dependent on extractive funding, have largely closed. Each new area of digital production that could have become a commons has been captured before the community could organise.

The Political Question

The commons is not finished. The everyday practice of sharing, contributing, and building continues across thousands of projects and communities. But the transformative potential, the idea that the digital commons would be a powerful force for expanding participation and improving social justice, has been largely neutralised. The commons now functions mostly as infrastructure for a system that has absorbed it.

What changes this is not better technology. It is political choice: about who governs shared digital infrastructure, where value flows, and who makes the rules. Some governments are beginning to make those choices at scale. This series examines what that looks like in practice, and what Australia stands to lose by not making them.

Part one of three. Next: : what France and Germany are actually doing about digital sovereignty, and what the €264 billion number means. Related reading: Unicorns Build Monocultures on Germany’s Sovereign Tech Agency and the NLnet Commons Fund. Don’t Let the Asphalt Bury the Garden on AI and the open-source question from an earlier angle.

Sources

The commons tradition and its decline

• Stalder, F. (2026). The decline of the digital commons. Theory, Culture & Society. DOI: 10.1177/02632764261432877. The theoretical framework this post draws on throughout; the three-part enclosure analysis (fix, pre-competitive cooperation, enclosure via cloud/SaaS) is developed in detail here
• Ostrom, E. (1990). Governing the Commons. Cambridge University Press: the foundational empirical refutation of Hardin’s “tragedy of the commons”
• Benkler, Y. (2006). The Wealth of Networks. Yale University Press: commons-based peer production as a third production paradigm beyond market and bureaucracy

Enclosure: legal and platform mechanisms

• Lametti, D. (2012). The cloud: Boundless digital potential or enclosure 3.0? SSRN. DOI: 10.2139/ssrn.2077742. On cloud computing as a mechanism for circumventing copyleft
• Zuboff, S. (2018). The Age of Surveillance Capitalism. PublicAffairs: on the enclosure of social commoning through data extraction
• Stalder (2026), pp. 10–11: Google/Android and the social media enclosure of cultural commoning

AI and the commons

• European Commission Joint Research Centre, Workshop on Generative AI and the Future of the Digital Commons, March 2025: the Wikimedia representative’s statement about AI slop and quality control (cited in Stalder, 2026, p. 13)
• TollBit (2025). AI scraping is on the rise: State of the bots Q4 2024: click-through rate reduction of up to 97% from AI-generated answers
• Mariani, R. (2023). The dead internet to come. The New Atlantis, 73: 34–42. On AI-driven simulation replacing social engagement
• Walter, Y. (2025). Artificial influencers and the dead internet theory. AI & Society, 40(1): 239–240

Pre-competitive cooperation

• Linux Foundation, linuxfoundation.org: the organisation and its funding model
• Stalder (2026), pp. 9–10: the analysis of commons as pre-competitive cooperation mechanism

• Digital Commons
• Open Source
• Ai
• Enclosure
• Platform Capitalism
• Digital Sovereignty
• Commons Governance