I.
The Frontier Lab is the institutional framework through which the future is shaped today. It is the technologists who are shaping this future, and they do so on behalf of and in the interests of private companies such as OpenAI or Anthropic, which are sustained by investments on a scale such that the collapse of these companies would result in an economic—and thus societal—catastrophe.
What is missing is the democratic counterpart to this model. While the world emerging from the Frontier Labs is radically digital, the system we use to shape society and politics is radically analog. The 21st century meets the 19th century. It is clear who wins here.
The logical conclusion, if we want to save democracy by reinventing it, is this: the institutions of tomorrow must be shaped in accordance with the possibilities of tomorrow. This applies just as much to public administration as it does to schools, universities, political institutions, parliaments, political parties, and the political arena. So far, we have seen this: the technologists are creating facts on the ground, while the democrats are watching.
When it comes to the future of democracy, it is therefore crucial to place institutional innovation at the center of our thinking: What form of institutions is appropriate for the age of AI, in which the foundations of politics and society are undergoing fundamental change? And what can we learn from those who are already shaping the institutions of the future—albeit without democratic impetus?
It is always helpful to listen to someone like Sam Altman, CEO of OpenAI, or Dario Amodei of Anthropic when they speak about the implications and consequences of their work—especially the implications for democratic practice, the future of work, or social cohesion —unlike the intellectually ambitious Amodei, the profoundly clueless Altman makes it deeply unsettling when one makes the mistake of expecting these technologists to provide the answers to tomorrow’s democratic questions.
What they have delivered so far is infrastructure, not political theory; the task is to democratize this infrastructure and create new infrastructures that, in turn, transform democracy.
In a present that has lost all balance, it is a recurring pattern to turn to precisely those for solutions who are responsible for the problems in the first place. It is not the technologists who should develop the democracy of tomorrow—they lack the ambition or training for this, and it is neither in their interest nor their mandate—rather, the ideas should come from politics and civil society, which, however, often lack the technological understanding of the challenges and opportunities presented by, for example, large language models.
This fundamental contradiction must be resolved. Genuine answers to the democratic crisis will only be found if non-technologists acknowledge their own lack of knowledge and strive to understand the technological realities of our time—not in pursuit of technological or technocratic solutions, but for solutions that are on par with the technology itself.
II.
It is the radical nature of their form that makes Frontier Labs so unique. They are free from democratic oversight, yet they act—in terms of economic power and institutional disruption—almost like states. Their intentions are opaque, yet they are treated as partners, as rational actors in a game with rules. At the same time, they remain in a constant state of experimentation, obeying few rules other than those of their own internal logic and venture capital.
OpenAI, Anthropic, and Google are thus replacing essential institutions, structures, and ways of thinking in political practice. We are witnessing societies around the world being transformed into experimental spaces, without the parameters of these experiments being clear, without the goals and assumptions being communicated in an understandable way, and without any effective means of public participation or political oversight.
The way Frontier Labs operate is, in many respects, a mystery, and former employees of OpenAI or Anthropic are contractually bound to silence. But there are reports, such as that by Calvin French-Owen, from which a few principles applicable to OpenAI can be derived; I believe these also offer insights into the form that future democratic political practice might take.
First, according to French-Owen, OpenAI is “incredibly bottoms-up, especially in research”: “When I first showed up there, I asked about the roadmap for the next quarter. The answer I received was, ‘There isn’t one’ (although one does now exist).” Good ideas can come from anywhere, and it’s often not really clear in advance which ideas will prove to be the most fruitful. Instead of a large-scale ‘master plan,’ progress is iterative and only becomes apparent when new research findings bear fruit.”
Second, thanks to its bottom-up culture, OpenAI is highly meritocratic. “In the past, leaders within the company were promoted primarily based on their ability to develop good ideas and then implement them. Many leaders who were incredibly competent weren’t particularly good at things like giving presentations to the entire staff or political maneuvering. That plays a lesser role at OpenAI than it might at other companies. The best ideas usually prevail.”
Owen-French describes “a strong bias toward action (you can just do things)” as the third key aspect of OpenAI’s culture: It’s not uncommon for similar but independent teams to come up with different ideas. What’s always important is that projects are tackled by a small handful of people without them having to ask for permission. “As soon as these projects prove promising, teams quickly form around them.”
A fourth element of the Frontier approach is the researchers, who act as their own “mini-executives.” “There’s a strong tendency to work on one’s own project and wait to see how it develops. This leads to a logical consequence: Most research arises when a researcher is specifically assigned to a particular problem. If something is considered boring or ‘solved,’ it’s unlikely anyone will work on it.”
For Owen-French, a fifth factor in OpenAI’s success is the ability to change course quickly: “It’s much better to do the right thing as soon as you receive new information than to stick to a course just because you had a plan. It’s remarkable that a company as large as OpenAI still maintains this attitude—Google obviously does not. The company makes decisions quickly, and when it chooses a direction, it commits fully to it.”
Bottom-up, flat hierarchies, performance over management, autonomy and trust, action and experimentation, and the ability to overturn everything: A good summary of the Labs’ principles is provided by Nathan Lambert’s blog post, which boils down to the statement: “you ship your org chart”—the product is ultimately a consequence of decisions that are embedded in and reflect the company’s organizational structure.
The unifying idea behind this: These are small units operating here in the spirit of a larger plan, and the pressure—the pressure to justify—lies not with the new to prove what it can do, but with the old to show why we should keep it.
So if one wants to develop a new democratic practice, it’s not about breaking the old system at any cost; it’s about recognizing the possibilities—and also the necessities—that emerge in the practice of Frontier Labs and how they can be harnessed democratically.
For, of course, the Frontier Labs themselves, in their current form, are not the building blocks of a democratic future. In her excellent book *Empire of AI*, Karen Hao describes the inner workings of OpenAI with critical clarity—in the company’s early days, she writes, about a quarter of the employees were women or non-binary individuals; the executives were almost all male and white.
And the ethos of AI companies—as articulated, for example, in the texts by Dario Amodei, CEO of Anthropic, and in the widely read manifesto by Leopold Aschenbrenner, who was part of OpenAI’s “Superalignment” team until 2024—is quite clear: It is a small group of people who have the right to steer the development of AI.
The logic and strategy are thus Leninist: “There are probably only dozens of people who truly ‘need to know’ the key implementation details for a given algorithmic breakthrough at a given lab,” writes Aschenbrenner, “you can vet, silo, and intensively monitor these people, in addition to radically upgraded infosec. ”
Or as Lenin put it in 1920 in “What Is To Be Done?”: “The organization of the revolutionaries, on the other hand, must above all and primarily include people whose profession is revolutionary activity.” And further: “This organization need not necessarily be very extensive and must be as conspiratorial as possible.”
III.
Democracy works differently. But perhaps, at certain moments, it must function this way—at least in part. For the question is, after all, what is meant by democracy: form or content? The two complement each other in unique and not always identical ways.
We are currently experiencing a crisis of the form of democracy, which is linked to a crisis of content. It is not easy to say which comes first: the systemic contradictions of a style of governance that does not correspond to the possibilities, the requirements, or the experience of our time—or systemic resistance that is currently unleashing itself with full vehemence in a fascist manner.
But that is not the point at all: the issue is to see the problem—and thus also the solution—in the very structure of the system itself—understood not as an ideology, but as infrastructure. So how is information processed today—socially and, above all, politically? How are decisions made, communicated, and implemented?
What lies ahead is a complete overhaul of the system, in which individual parts can be retained. This requires courage and determination, and likely structures similar to those found in the Frontier Labs—the democratic contradictions inherent in this form of Frontier Politics must be articulated and resolved. The practice itself should be put to the test.
This would not mean running a political party like a business or a startup; it would mean generating the energy necessary to arrive at experimental, new, and better answers and solutions to current problems.
All of this has very practical implications. Frontier Labs are drawing in money and talent that are lacking elsewhere. It is no longer a cliché that in organized politics, at many levels—from the very bottom to the very top—talent is lacking, ambition is misdirected, and promising people are held back. There are so many misguided incentives here that politics, in its current form, will never win the competition for the best minds.
These minds, in turn, do not follow the same logic as political parties, which date back to a time when linear résumés and predictable careers were the norm; therefore, the goal cannot be to replace one or two generations of professional politicians with a new generation of professional politicians.
The logic of new technological possibilities would render the profession of the career politician obsolete. Politics as a profession, as Max Weber called it, is a product of the first industrial revolution of the 19th century; under the conditions of the coming industrial revolution, this field, like virtually all other areas of society and institutions, will also change.
These changes generate fear and centrifugal forces. The forces of resistance from the old system are strong, especially because they affect people who have found their role, purpose, and security within that system.
All of this can now lead to a turning away from technology itself—the infrastructure that underpins our society, economy, and, consequently, politics. Or one can turn toward these technologies and try to understand and shape them, attempting to see and harness their inherent and new logic to create institutions that are suited to the conditions of the times for which they are intended.
Of all people, Bill Gates has just summed it up: We need a plan for how we want to deal with AI. One could also say: We need politics. We need politics that operates at the level dictated by technology. We need new politics.
What we’ve seen so far, in various forms, is the organized lack of ambition on the part of those in power. The disruptions of our time, however, are real and cannot be addressed with the usual mediocrity.
