{"slug": "put-ai-under-social-ownership-for-social-purposes", "title": "Put AI Under Social Ownership for Social Purposes", "summary": "A Jacobin essay argues that artificial intelligence should be placed under social ownership to serve public purposes, proposing purpose trusts as a more pluralistic alternative to government control. The piece critiques a Bernie Sanders proposal for a $7 trillion AI sovereign wealth fund and cites concerns about political power imbalances, while advocating for sharing AI's wealth and incorporating public regulation within AI enterprises.", "body_md": "# Put AI Under Social Ownership for Social Purposes\n\nPublic ownership may seem like the most obvious solution to the problems posed by AI. But purpose trusts could offer a more pluralistic form of social ownership — one that subordinates profit to a variety of public purposes.\n\nA lively debate has been taking place in Jacobin over what to do about artificial intelligence. AI is clearly a technology with potential for enormous benefit and harm — and perhaps both at the same time. Financially, it could put even more income in the hands of billionaires, or it could produce a bubble the implosion of which pushes the net worth of investors, innocent and guilty alike, off a cliff. Either way, it needs to be redirected away from narrow financial imperatives and toward social ends, and this requires a change in ownership.\n\nBut what change, and how? The obvious go-to for socialists is to give government a controlling interest, and that’s what Bernie Sanders has [proposed](https://www.sanders.senate.gov/press-releases/news-sanders-introduces-legislation-to-create-7-trillion-ai-sovereign-wealth-fund/): use a onetime tax to transfer half the equity in US-based AI companies to a new sovereign wealth fund, which would then have veto power over future decisions affecting the technology while sharing in the profits it generates. But there is also unease with the idea, not least because Donald Trump has also floated federal [ownership stakes](https://www.reuters.com/business/trump-says-his-team-will-look-into-us-taking-stake-ai-companies-2026-06-05/) — and even a remote similarity is grounds for suspicion. More precisely, given the state of American democracy and the grotesque imbalance of political power favoring the ultrarich, it is reasonable to doubt that a large public stake would yield the outcomes we want to see from it.\n\nTo design social ownership so it truly reflects social purposes in a globalizing world, we need to rethink the issue from the ground up. Just why do we need to remove ownership and control from the hands of wealthy investors? Exactly what changes are we trying to bring about?\n\n## The *Why* and *How* of Public Ownership\n\nThe first question has a three-part answer. First, we are suffering from the overbearing power of an oligarchy that transforms vastly excessive wealth into suffocating political and social influence. If AI generates the economic benefits advertised for it, even greater riches will follow. AI companies convert vast stores of collectively produced knowledge and culture, along with publicly supported research, into private wealth. If AI does produce the economic benefits its exponents promise, those gains should be broadly shared rather than accrue to a small group of plutocrats. Second, the breakneck pursuit of market domination is generating immense costs while failing to deliver benefits that can’t be readily marketized. It is undertaken either nakedly for enrichment or in the name of a perverse, zero-sum notion of patriotism (any loss of market share to Chinese companies, some say, imperils America’s “national security”). Finally, AI is a complex, rapidly evolving technology, and, as such, it is almost impossible to regulate.\n\nYes, broad regulations on the use of AI and the resources it vacuums up are feasible and should be enacted as soon as possible. Personally, I’d begin with a prohibition on any chatbot or similar program using the first person in its messaging: “I” is a lie. But regulation of how models are trained, how they respond to prompts, and how they may soon be integrated with reasoning engines — these matters are almost impossible to regulate from the outside. It’s a game of never-ending catch-up that regulators, even if aggressive and well-motived, are unlikely to win.\n\nSo, to sum up the *why*, we need to share the wealth, dial back the profit motive, and incorporate public regulation within the enterprises developing and applying AI.\n\nWhich brings us to the crucial problem of *how*. Public ownership would deal with the first problem, and no doubt that’s a major reason why Sanders and others want to take this step. It might also help with the third, since, if nothing else, the government as equity investor could demand that all information relevant to regulation be made public in real time. Those are both significant advantages that should not be disparaged.\n\nBut public ownership also poses problems of its own. One, as a previous Jacobin writer correctly [emphasized](https://jacobin.com/2026/07/ai-democratization-working-class-institutions), is that government control and democratic control are typically quite far apart. Even in the best of circumstances, people get to vote only periodically for political representatives. By the time decision-making snakes its way down to lower levels in the bureaucracy the democratic mandate is weak. And today those circumstances are light years from democratic perfection.\n\n## Public Pluralities\n\nYet the problem with public ownership is even more fundamental. To see it, we need to think concretely about what it is we want AI to do — and not do — and ask how ownership models alone can resolve potential conflicts. Here are several issue areas that occur to me:\n\n*Privacy and personal freedom*. AI is being deployed to create systems of private and public surveillance without equal in human history. This ranges from surveillance pricing, where what we pay is a function of what an AI agent determines we are willing to pay, to state profiling of our political values for purposes of social control. A frightening prospect is the use of AI to target and annihilate places and people beyond human oversight, as exemplified by Israel’s Lavender program.*Management and employment*. Much of the putative economic value of AI stems from its claimed ability, now or soon, to assume direct management of businesses, government agencies, and other organizations, with the wholesale dismissal of the human beings who now do this work. That can lead to frustratingly unresponsive or unhelpful services — an aspect of Cory Doctorow’s[enshittification](https://www.newyorker.com/culture/infinite-scroll/the-age-of-enshittification)— and of course it raises the specter of mass unemployment. Yet replacing drudgery with automation can also be an enormous social boon; just think of the millions of secretaries whose typing, filing, and other repetitive clerical chores were taken over by word processing apps. (I was in their ranks once.) Key choices about how organizations should be run, and who will be needed to run them, are at stake.*Environment and resources*. Unlike earlier iterations of digital technology, AI makes massive, almost mind-boggling demands on our water resources, energy systems, and rare minerals. This is due to its brute-force training methods, and it might be mitigated by more specialized training combined with greater use of logic. In any case, there are important issues to resolve about how to trade off computing capacity and resource use, along with whose resources should be tapped and how they should be compensated for them.*Applications across society*. As much as attention has focused on the issues above, this may be just as important in the long run. Already educators are struggling with the slash-and-burn effects of AI on classrooms. The impacts on writing, music, and visual culture could be enormously positive, negative, or (most likely) both. AI shows great promise in research, but a lot will depend on whether it augments our understanding of the fields to which we apply it, or simply pulls results out of a black box. This would matter even if agents were perfectly reliable, which they’re not. Who can we trust to design these applications in a way that expands our potential as learners and creators?\n\nThere are important conclusions to draw from even this cursory survey. One is that many kinds of choices must be made. Some are distributive — who gains, who loses — but others involve fundamental values. Another is that, given the many dimensions of AI’s future development, it’s impossible to specify a single public interest that all of us would agree to sign onto. It’s difficult to overstate how important it is to understand this. The public whose interests should guide us is the actually existing one, with all its diversity of attitudes and priorities — not an imaginary public in which everyone thinks just the way you or I do.\n\nEven if everyone in the United States adopted progressive values, choices would still have to be made among competing priorities. The issue is not data centers everywhere versus nowhere, or AI-generated documents versus only human-written ones, or any other dichotomous choice. It is about priorities, emphases, adjustments, and reimagined designs. The current choices are mercenary and terrible, but there is no simple alternative upon which everyone will agree.\n\n## Purposeful Ownership\n\nSo we need not only democracy in the AI world but also true pluralism: different implementations existing side by side from which we can choose. This is also necessary for standard economic reasons, since we won’t know in advance what approaches to the development of AI will turn out to be best, either in technical or social terms. Again, pluralism.\n\nFortunately, we already have the beginnings of the alternative we need: ownership by purpose trusts. Up to now, this has progressed one retirement at a time, as public-minded owners such as Patagonia’s Yvon Chouinard, who chose to place control of their companies in the hands of trustees legally bound to protect particular purposes. In the case of Patagonia, this was an [environmental commitment](https://www.patagonia.com/ownership/).\n\nUnder the stronger version of the model proposed here, profit would become a means, not an end. The trust deed would stipulate that earnings, after necessary reinvestments, be devoted to advancing the firm’s overriding purposes rather than distributed to private investors. If you scan a directory of such firms, you’ll find a variety of mandatory purposes, ranging from worker benefit to community development to enhancing local food systems, with many designating more than one of them. Which is how it should be: we are a diverse society, and our purposes are many and do not always lead to the same choices.\n\nHowever, to adapt this model to AI we need to make two very large changes. First, the public, acting through its elected representatives, would have to initiate the conversion of existing firms to purpose trusts. This would require a process for recruiting trustees and, above all, financing the purchase of firms’ equity. Some form of tax on profits would obviously be essential. Second, the purposes would need to be chosen not by enlightened, aging entrepreneurs but by society itself. Because multiple trusts could coexist, they could pursue a range of purposes that would not always agree with one another. Some roughly democratic method would be needed to ensure that the allocation of funding among the trusts corresponds to the proportion of the community that supports respective purposes.\n\nA side note: Anthropic is not itself a trust, but its governance structure includes a long-term [benefit trust](https://www.anthropic.com/news/the-long-term-benefit-trust). It differs from the mode described here in several crucial ways: investors maintain their claims on profits; Anthropic’s mandate was focused on the long-term risks posed by AI, leaving most business questions unaddressed. It is relatively easy for equity owners to dissolve the trust whenever they wish. It is no secret that Anthropic is preparing for an initial public offering.\n\n## The Purpose-Trust Advantage\n\nAnother big advantage of the purpose-trust approach over public ownership is that it is far better suited to handle the pushes and pulls of the global economy. It can partner with foreign firms if it chooses. It can pursue international strategies consistent with its purposes separate from the passing predilections of politicians — a huge consideration since China is also a key player. And it can use these relationships to promote people-serving purposes beyond our borders, which can add a potentially progressive twist to globalization. This flexibility follows from the model: once capitalized, trust-owned firms are operationally autonomous and governed by their legally binding purposes.\n\nThis last point is key. Consider a trust-owned AI firm whose legally binding objective is to help users be more creative and productive. This would obligate the firm to develop tools that enhance user’s capacities and to prohibit the development of applications that are routinizing or disempowering. But such a broad purpose could only be realized through sustained interaction with users. Some of this would take the form of standard market feedback, as when a gardener buys an AI helper that makes gardening more worthwhile, or someone learning a new language purchases an AI teaching tool. Our hypothetical trust-owned business would still have to earn enough revenue to cover its costs, just like everyone else.\n\nNow imagine that its tools were intended for use by people at work, making it possible for unions and professional or advocacy groups to have a voice in their design and deployment. Organizations of all sorts could partner with our AI developer, since trust ownership is a form of social ownership that preserves the firm’s independence. One trust-owned firm could also collaborate with another when their missions overlap, or if they operate in related parts of the economy — for example, an AI developer joining forces with a pharmaceutical company, with both owned by purpose trusts. And, as mentioned above, national borders do not set boundaries on these kinds of partnerships unless specific laws are passed to regulate them.\n\nConversely, two trust-owned firms with opposed missions might compete if they’re in the same market — not to amass the greatest possible profit but to advance their respective missions. This would be social ownership without the myth of a homogenous, same-thinking society.\n\nOf course, unlike public ownership, social ownership via purpose trusts does not come ready-made. We have limited experience to draw on, and important questions remain unanswered. There will surely be a learning curve. But the need for a vibrant, diverse, and purpose-driven alternative economy is more urgent than ever, and the way to figure it out is to start creating it.", "url": "https://wpnews.pro/news/put-ai-under-social-ownership-for-social-purposes", "canonical_source": "https://jacobin.com/2026/08/tech-ai-purpose-trusts-public-interest", "published_at": "2026-08-25 16:56:14+00:00", "updated_at": "2026-08-25 17:15:31.540275+00:00", "lang": "en", "topics": ["ai-policy", "ai-ethics"], "entities": ["Jacobin", "Bernie Sanders", "Donald Trump"], "alternates": {"html": "https://wpnews.pro/news/put-ai-under-social-ownership-for-social-purposes", "markdown": "https://wpnews.pro/news/put-ai-under-social-ownership-for-social-purposes.md", "text": "https://wpnews.pro/news/put-ai-under-social-ownership-for-social-purposes.txt", "jsonld": "https://wpnews.pro/news/put-ai-under-social-ownership-for-social-purposes.jsonld"}}