Something strange is happening with intellectual property. For decades, IP regimes have served as the primary mechanism through which creative and intellectual labor is governed, compensated, and enclosed. In the last few years, the largest technology companies in the world have trained generative AI systems on vast corpora of copyrighted material, and the legal and economic frameworks that were supposed to prevent this have proven largely unable to respond. As a reaction, progressive sections have rallied behind intellectual property as a way to stem the untrammelled expropriation of people’s work. After decades of battles against intellectual property, it is now being latched onto as a practical, sometimes moral, lifeboat. Conversely, the sidelining of the IP regime by those who used to support it vigorously earlier – especially agents of large capital – is not surprising, but it is still jarring.
After decades of battles against intellectual property, it is now being latched onto as a practical, sometimes moral, lifeboat.
It is worth acknowledging this strange role reversal and the rapid technological changes that are occurring. We are in a period where the shape of the AI market, and the market for its outputs, are both not clear. Copyright claims have been filed against companies that, in some cases, are already becoming irrelevant. The pace of change is outstripping the pace of legal response. And the conversation about what is being lost tends to focus on theft from artists, which is real and important, but which also obscures a broader dynamic: the displacement of the labor of art itself. When AI models output images and video, even when this output is derivative and replicative, this replication is carried out by workers in the AI supply chain: workers in quarries, mines, chip factories, and data centres.
And yet there is almost no major discussion about “stealing” from software engineers. The conversation is instead framed around unemployment and labor displacement.
Consider the contrast with software. In the world of software development, AI-assisted coding tools are rapidly automating large portions of what programmers do. And yet there is almost no major discussion about “stealing” from software engineers. The conversation is instead framed around unemployment and labor displacement. This asymmetry is revealing. It suggests that parts of the cultural sector’s attachment to IP as the primary frame for understanding creative labor has made it harder to see what is actually happening: control over the labor process, a reorganization of who does intellectual and artistic work, under what conditions, and for whose benefit. The adage of software wanting to be free turns out to be a preview of a logic that is now being crudely extended to all forms of intellectual production. Is it only software that wants to be free – or were we wrong to believe that software should be free? Two facts are now in sharp relief: intellectual and artistic production are always a result of collective labor, and a great deal of conscious effort goes into actively designating intellectual and artistic products as property. There is nothing natural or ahistorical about this property designation.
When a new technological capacity emerges, the practices that develop in the absence of regulation tend to become sticky. They create path dependencies, economic interests, and political constituencies that make subsequent regulation much harder.
This is, of course, cold comfort to artists who might lose their livelihoods. How did we get into this situation? Part of the answer lies in the nature of legal vacuums that were created because we did not act fast enough. The law cannot anticipate every new technology. When a new technological capacity emerges, the practices that develop in the absence of regulation tend to become sticky. They create path dependencies, economic interests, and political constituencies that make subsequent regulation much harder. Letting the market figure it out before intervening is therefore not a neutral stance. It is a stance that favors whoever moves first and fastest, which, in the case of AI, means a small number of very large, very well-capitalized firms. By the time the market has figured it out, the defaults are already set. Changing those defaults can be really costly. In the case of AI, ruling now that AI models cannot be made because they steal copyrighted materials would crash stock markets, which is precisely why such a ruling will not be allowed to be made. Acting early is sometimes necessary.
In the case of AI, ruling now that AI models cannot be made because they steal copyrighted materials would crash stock markets, which is precisely why such a ruling will not be allowed to be made.
This dynamic has a direct precedent in the platform economy. The monopolization of the broader digital economy, I would venture to argue, began with the monopolization of art and culture. First, platforms became the dominant distribution channels for creative work: simply talking to an audience of friends turned into work. Then, creators on those platforms became influencers. Influencers are now a major marketing channel, and platforms have become, in many cases, almost the only viable markets. Each step seemed like an expansion of opportunity for individual creators, and at each step, the strength of platforms grew. The individualization of art and culture through platforms was presented as liberation, as democratization. The monopolization of art was the template for the monopolization of everything else.
The current moment is actually a good opportunity to create new regimes for creativity that are not IP-based. The philosophical justification for collective rights over intellectual production has never been stronger, and the practical need for institutional forms that recognize this has never been more urgent.
This suggests that the question we should be asking is: what is the appropriate form of social governance for what is clearly social production? What is our equivalent of ground rent? For Thomas Paine, ground rent was the mechanism by which people captured a share of the value generated by forcibly enclosed land. As I have written elsewhere, AI firms are now enclosing economically relevant intellectual capacity. The question is whether we can develop institutional forms that treat this capacity as a collective resource rather than a private one.
As I have written elsewhere, AI firms are now enclosing economically relevant intellectual capacity. The question is whether we can develop institutional forms that treat this capacity as a collective resource rather than a private one.
If the frame is IP, then the policy responses are familiar: stronger copyright enforcement, licensing regimes, maybe some form of compulsory licensing for AI training data. If the frame is instead the public domain and collective rights over intellectual production, then a different set of possibilities opens up. We can talk about public AI infrastructure that is not controlled by private firms, or collective bargaining structures for creative workers that are not dependent on individual IP claims.
We should also be conscious of and respond to the relationship between capital accumulation and speed, which is, in the end, the core problem. The firms that are enclosing the intellectual commons are able to do so because they can move faster than institutions can respond, because large amounts of capital make it a trivial matter to move fast.
Finally, it is worth noting that statistical systems of the kind that currently dominate the AI landscape are not the only way to build intelligent tools. They are, however, the way that is most amenable to the logic of enclosure, because they require massive datasets and massive compute, both of which favor concentration. Looking beyond deep learning is also part of a public domain framing for AI and development.
Looking beyond deep learning is also part of a public domain framing for AI and development.
We find ourselves in a crisis that brings clarity. AI has made visible the deep tensions in how we organize intellectual production, tensions that the IP regime has papered over for decades. If we respond only by trying to shore up that regime, we will miss the opportunity to build something better. The task is to take the collective nature of intellectual production seriously and to build institutions that reflect that reality, before the defaults are set even more and the window closes completely.
The series brings together expert voices and was commissioned to inform the development of the issue brief by IT for Change, ‘Governing AI for the Cultural Commons: Beyond Intellectual Property’, under the AI, Culture and Intellectual Property Subgroup of the UNESCO Global Civil Society Organizations (CSO) and Academic Network on AI Ethics and Policy.
New article every Wednesday! Watch this space for more thinkpieces and read the issue brief here.
