Jill Lepore says Silicon Valley is building an ‘artificial state’ — and repeating old political fantasies

Historian Jill Lepore says Silicon Valley’s biggest companies are not just making software — they are slowly adopting the role of government. Her new book, The Rise and Fall of the Artificial State, argues that the tech industry has spent years presenting itself as a superior form of public authority, with artificial intelligence now becoming the latest vehicle for that ambition.

In a conversation on TechCrunch’s Equity podcast, Lepore traced that mindset from social platforms that distort civic life to AI leaders who talk openly about making government more efficient, more responsive, or even unnecessary. Her central warning is blunt: many of the people selling the future of AI appear to be rebranding an old political fantasy, not inventing a genuinely new one.

The discussion comes at a moment when AI companies are drawing closer to public institutions, public messaging and public power. From Anthropic’s “constitution” for Claude to Sam Altman’s musings about an “AI president,” Lepore sees a familiar pattern: technology leaders using utopian language while presenting their products as tools that can substitute for democratic institutions rather than serve them.

What is the “artificial state”?

The “artificial state” is Lepore’s term for the way technology platforms can begin to function like quasi-governments, even if they are privately owned and accountable to shareholders rather than voters. In her view, this is not only about companies influencing politics. It is about platforms quietly taking over tasks that once belonged to democratic institutions, such as shaping public discourse, distributing information and defining what people see as reality.

Lepore argues that this shift has been building for years. Twitter, she notes, was often described in civic terms — famously as a kind of “town hall in your pocket” — even though it was never a public square in any legal or democratic sense. Facebook and similar platforms altered how communities received news, which voices gained prominence and how local institutions were bypassed.

Her broader point is that the rhetoric of disruption often obscures a more consequential change: tech firms do not simply make tools. They increasingly manage social behavior, mediate collective life and, in some cases, decide what counts as legitimate participation in the public sphere.

Why does Lepore think Silicon Valley talks like government?

According to Lepore, the language is not accidental. She sees a pattern in which tech executives borrow the vocabulary of states, constitutions, rights and citizenship because it grants their products moral weight and historical importance. If a company can frame its platform as a better way to organize society, it can also justify enormous influence over people’s lives.

That framing becomes especially potent in AI. Instead of selling software as a productivity tool, AI companies increasingly describe their systems as assistants, advisers, judges, agents or even stand-ins for human decision-making. Lepore’s concern is that this language encourages users to think about machines as institutions rather than products.

She argues that the result is a subtle political shift. A company that controls a widely used AI model is not merely shipping code; it may be shaping what answers people trust, which perspectives are amplified and which decisions are treated as neutral or objective.

Lepore’s argument is that Silicon Valley has a habit of describing products in civic language while steadily building private systems that behave like public power.

How AI fits into Lepore’s argument

AI is the newest and perhaps most ambitious version of this trend. Lepore sees today’s AI leaders as reviving a long-running Silicon Valley dream: that better technology can make messy democratic governance obsolete. In that story, automation does not just improve public administration — it replaces the human and political processes that make government accountable in the first place.

That is why she is skeptical when AI executives suggest that their systems might eventually help run public services, resolve bureaucratic bottlenecks or make government “smarter.” To Lepore, such proposals are often presented as efficiency upgrades but can hide a larger ideological claim: that democratic institutions are too slow, too flawed or too compromised to deserve much confidence.

Her critique does not stop at the software. She is also challenging the worldview behind it — the belief that technical design can solve political conflict, and that governance is simply another optimization problem.

What examples does she see as warning signs?

Lepore points to several highly visible examples as evidence that the industry is comfortable blurring the line between product design and political philosophy. Anthropic’s attempt to give Claude a “constitution” is one. Altman’s talk about an “AI president” is another. Even the grandiose way companies market AI tools, she suggests, reveals how deeply the industry has absorbed the idea that software can model or replace public authority.

Those examples may sound playful or speculative, but Lepore treats them as important signals. In her reading, the point is not whether an AI system literally becomes president. The deeper issue is that the language normalizes the idea that governance can be delegated to technical systems designed and controlled by private actors.

She argues that this should worry anyone who cares about democratic legitimacy. A public institution can be criticized, voted out or reformed through law. A platform or model can be updated behind closed doors, with users offered little more than a terms-of-service explanation.

Why does she keep returning to science fiction?

Lepore says science fiction has long offered Silicon Valley a script for imagining the future, but often the wrong parts of that script get borrowed. One text she highlights is E.M. Forster’s 1909 story The Machine Stops, a prescient warning about isolated people living inside a world run by a powerful system they no longer understand.

For Lepore, the story is a useful lens because it captures the emotional and political logic of a system that feels omnipresent yet fundamentally fragile. She argues that some of today’s AI rhetoric sounds less like careful planning and more like a romance with collapse: a fantasy of total dependence on machines combined with the assumption that human institutions will naturally fade away.

Her criticism extends to the industry’s reading habits as much as to its products. In her view, Silicon Valley’s leaders often do not engage seriously with the cautionary tradition of science fiction. They borrow its imagery but miss its skepticism.

How does Forster’s story relate to modern AI?

Forster’s work remains relevant, Lepore suggests, because it imagines people surrendering agency to a system they praise even as it isolates them. That, she argues, is a recurring pattern in the tech industry’s self-description: dependence on the machine is portrayed as progress, while the loss of human judgment is treated as a side effect rather than a crisis.

Lepore’s memorable comparison, as discussed in the podcast, casts the story in modern terms — not as a sleek futurist prophecy, but as an eerily familiar portrait of digital self-absorption. The point is that the warning has not gone out of date just because the devices are smaller and faster.

The political history behind today’s AI hype

Lepore’s book reaches beyond the current AI boom. She places the present inside a longer American story about technology, power and democratic weakness. The idea that a new communications system will make society more rational or more free is not unique to the age of large language models. It has appeared repeatedly, from the early era of personal computing through the platform age and now into AI.

That historical perspective is part of what makes her critique more than a reaction to the latest product launch. She is arguing that the same underlying temptation keeps resurfacing: the belief that a technical system can do what politics cannot, or should not, do.

In that framework, the most dangerous promise is not convenience. It is replacement. When companies imply that AI can substitute for public deliberation, administrative judgment or even leadership itself, they move from innovation into constitutional imagination.

What does Lepore think is being replaced?

Lepore says the institutions under pressure are the basic building blocks of democratic life: local news, public agencies, shared civic spaces and the idea that people collectively govern themselves through imperfect but accountable systems. If those functions migrate into private platforms, the public may lose not just efficiency but visibility into how decisions are made.

That concern has grown more urgent as AI systems are embedded into search, workplace tools, consumer products and administrative workflows. The more normal it becomes to ask a private model for guidance, the less obvious it may be that important social functions have been privatized.

In Lepore’s telling, that is the essence of the artificial state. It does not announce itself as a takeover. It arrives through convenience, scale and flattering rhetoric.

Why Lepore says the artificial state is unsustainable

Lepore does not believe the artificial state can last forever. Her view is that systems which depend on private control over public functions eventually run into political limits, legal pushback or social distrust. The question is not whether there will be resistance, but how long it takes and what form it will take.

She suggests that one path to restraining the trend would involve renewed attention to democratic institutions: stronger public oversight, clearer rules for platform power and a willingness to treat AI not as destiny but as infrastructure that must be governed.

That is a more sober vision than the one often sold by the industry. Instead of replacing government, AI would need to operate within democratic constraints. Instead of assuming efficiency is always good, policymakers would have to ask who benefits, who is accountable and what gets lost when human institutions are displaced.

In Lepore’s view, the real issue is not whether AI can be made smarter, but whether society can stop private systems from quietly taking over public roles.

Why this debate matters now

The conversation arrives at a critical point for AI policy and public trust. Companies are racing to insert AI into consumer services, work software and public-sector tools at the same time that citizens are already struggling with misinformation, platform consolidation and declining confidence in institutions. That makes the question Lepore raises especially relevant: when does a useful technology become a governing power?

Her answer is that the line is crossed earlier than many people realize. Once platforms begin shaping civic reality, distributing information at scale and framing themselves as the remedy to politics, they are no longer operating at the edge of democracy. They are inside it.

For readers tracking the future of AI, the book and podcast appearance offer a useful corrective to the industry’s most ambitious claims. The discussion is not about whether machines are impressive. It is about whether the people building them understand the political consequences of describing them as replacements for institutions that took centuries to build.

Key points from Lepore’s argument

• Tech companies increasingly present themselves in civic or governmental language.
• AI systems are now being marketed as tools that could improve or even replace parts of government.
• Lepore sees this as a continuation of Silicon Valley’s long-running utopian messaging, not a break from it.
• Science fiction, especially cautionary works like The Machine Stops, helps explain the logic behind the industry’s ambitions.
• She believes the artificial state will ultimately face political and institutional limits.

What comes next for the debate over AI and power?

The broader fight is likely to center on accountability. If AI models become embedded in public life, then lawmakers, regulators, researchers and the public will need to decide what level of control private firms should have over information, decision-making and civic infrastructure.

Lepore’s intervention matters because it reframes the conversation. Instead of asking only whether AI is accurate, fast or useful, she asks whether it is becoming a substitute for democratic authority. That is a much larger question, and one that Silicon Valley’s most ambitious leaders seem eager to answer in their own favor.

For now, the debate over the “artificial state” is less about a single product than about a worldview. If the industry continues to describe its systems as better versions of government, the public may eventually have to decide whether those systems deserve trust — or whether they are simply private power wearing a civic mask.

That is the warning at the center of Lepore’s book, and it is what makes her critique resonate beyond the tech sector. In an age when AI companies are increasingly asking to be taken seriously as social institutions, she is asking an older question: who elected them?