Historian Jill Lepore — The Artificial State

Harvard historian Jill Lepore, whose earlier work "These Truths" became a defining single-volume American history, named Anthropic explicitly in her new book — out August 25, 2026 — as an example of a tech company quietly performing the functions of a government without any democratic mandate to do so. The book, "The Rise and Fall of the Artificial State," argues that the industry has built what amounts to a parallel political system, and it presents measurable evidence — not just philosophical complaint — for why that claim holds up under scrutiny.

Background You Need

For the past decade, the operating mythology of Silicon Valley has run something like this: technology companies democratize access, flatten hierarchies, and accelerate progress in ways that slow-moving governments cannot match. That story depends on the word "democratize" doing enormous work, and Lepore, who holds an endowed chair at Harvard and has been a staff writer at The New Yorker since 2005, has spent much of her career testing whether the words people use about power actually describe what power does.

The book grows out of a November 2024 New Yorker essay Lepore wrote under the same title. That essay landed in the middle of a presidential election cycle already warped by algorithmic platforms and AI-generated content, and it drew substantial attention outside academic circles. The expanded book, published jointly by Crown in the United States and Penguin in the United Kingdom, is positioned explicitly as a successor to Hannah Arendt's 1951 "The Origins of Totalitarianism" — a book that diagnosed how the machinery of modern mass society had rewritten the basic conditions of political life. Where Arendt was writing about bureaucratic states and industrialized violence, Lepore is writing about a new species of institution: the private company that governs without calling itself a government.

Jaron Lanier, whose own writing on technology has often taken an adversarial stance toward the industry's self-image, provided a blurb for the book calling Lepore "our best scholar of the culture of governance" and describing the AI-versus-democracy question as "the issue of our times." That endorsement matters because Lanier is not an outside critic — he has worked inside major tech institutions and knows the terrain. His framing of the book as a "voice of sanity" is a pointed signal about what kind of debate this enters.

What's New

The core argument Lepore makes is that the tech industry has built what she calls the "Artificial State" — a set of institutions and systems that perform governance functions (setting rules, adjudicating disputes, shaping public discourse, encoding moral values) while claiming to be doing something else entirely: running a marketplace, operating a platform, or simply shipping a product. The evidence she marshals is specific enough to be checked.

Anthropic, the AI safety company that built Claude, is named as an example of state-building in corporate disguise. The specific act Lepore points to is Anthropic's decision to hire a moral philosopher to write a "constitution" — a document that encodes values, sets behavioral limits, and defines what the system may or may not do. Lepore's argument is not that this was done in bad faith, but that it was done at all, by a private company, without any public process, consent, or accountability structure. Facebook's creation of its Oversight Board — internally referred to as a "Supreme Court" — is cited alongside it, as is Sam Altman's publicly stated view that an AI president might actually be a good idea. Taken together, these form a pattern: companies reaching for the vocabulary and the architecture of democratic governance while remaining answerable to no electorate.

The Twitter data point is where the book's argument becomes hardest to dismiss. Lepore documents that more than 90 percent of political tweets were produced by fewer than 10 percent of users during the period when politicians and journalists routinely described the platform as a window into what the public was thinking. That number does not describe a democratized public square — it describes a megaphone captured by an unusually active and unrepresentative minority. Any political operator or media strategist who used Twitter engagement as a proxy for voter sentiment was, by this data, working from a systematically distorted signal.

The book also takes on the industry's relationship with the science fiction that inspired it. E. M. Forster's 1909 story "The Machine Stops" — a tale about a civilization of isolated individuals connected only through a technological network, who eventually lose the capacity to survive when that network fails — predicts something far closer to the DoorDash-and-Zoom world of 2020 than to any utopia. Lepore's reading is that the tech founders who cite science fiction as a roadmap are often bad readers of the texts they claim: the stories they love most frequently end in catastrophe, not liberation. A separate observation in the book about Elon Musk's stated sci-fi favorites — that the politics embedded in those texts actively contradict the politics Musk espouses — lands as a specific, falsifiable claim about reading comprehension rather than a general complaint about ideology.

The structural villain in Lepore's historical account is the 1996 Telecommunications Act, which she identifies as the legislative moment when the rules governing media consolidation and platform accountability were written in ways that foreclosed later correction. She is careful to note this was not a conspiracy — it was the product of Newt Gingrich's Contract with America, a mainstream legislative agenda with broad support at the time — which actually makes the argument harder to refute than a simpler story about bad actors would be.

The Pushback

The obvious counterargument is that private companies have always built infrastructure that shapes public life — railroads, telegraph networks, telephone systems — and that regulatory frameworks eventually caught up. By this reading, Lepore is identifying a real lag in governance without proving that the lag is permanent or that the outcome is as dire as she suggests. The book's Penguin edition release date of August 25, 2026 means we are still weeks away from seeing the full academic response, and the most rigorous empirical challenges to her data will come from that quarter rather than from the industry she is critiquing.

Our read is that the Anthropic naming is the book's most strategically significant move. Lepore is not talking about Facebook circa 2016 or Twitter under its original management — she is talking about a company founded in 2021, with an explicit AI safety mission, by people who have thought carefully about institutional design. If her argument applies there, it applies everywhere. That is a harder claim to dismiss as nostalgia for a pre-tech world, and it is the one that developers working on AI systems in 2026 will need to be able to answer clearly.

Sources