Historian Jill Lepore has spent much of her career writing about the institutions that hold public life together. In her new book, The Rise and Fall of the Artificial State, she turns to a question that sits squarely inside the AI safety debate: What happens when functions once carried out by governments, communities and public institutions migrate to private technological systems?
Lepore is the David Woods Kemper ’41 Professor of American History at Harvard University, a professor at Harvard Law School and a staff writer at The New Yorker. Harvard lists her work as spanning history, law, technology and the politics of evidence.
I spoke with Lepore about why she sees ChatGPT as a historical turning point, how automation can hollow out public institutions even when it begins with good intentions, and why she thinks the central problem around AI is unchecked private power rather than technology itself.
The conversation has been edited and condensed for clarity. Answers have been lightly edited for length, repetition and false starts.
AI Safety Watch: You write that ChatGPT was your “Sputnik moment.” What changed for you?
Jill Lepore: To some degree, that’s a literary maneuver, to try to call attention to the nature of a turning point. I probably paid more attention to large language models than most humanists. I’m married to a computer scientist who was in graduate school in the ’90s working on hidden Markov models and natural-language processing, so it’s a field that I’ve paid a lot of attention to.
For most people around the world, it did come out of nowhere. The release generated a lot of hoopla, and part of that was because of the way OpenAI conducted the release. But the product itself really generated big conversations that we’re still in the midst of. I thought it was a useful inflection point around which to frame and introduce this book because I think a lot of people had a very strong reaction to it. There was kind of before ChatGPT and after ChatGPT.
What do you mean by the “artificial state”?
Lepore: I try to be careful in the book to say that it’s both a fantasy and a thing that is being built. I sometimes refer to it as half-built. What I’m trying to do is give a name to something that I think people haven’t been especially well able to name, and yet they have a sense of it.
What I mean by the artificial state is that increasingly the functions of governments have been taken over by, or have been outsourced to, private multinational corporations in ways that, just on the face of it, undermine the functions of a liberal democratic nation-state.
There are specific, overt ways in which these corporations have adopted the language of government and institutions that represent the consent of the governed and involve checks and balances, accountability, transparency and commitment to the public interest, none of which applies in the artificial state.
The term is really an umbrella phrase to call our attention to the ways in which the liberal democratic constitutional nation-state has been eroded by these forces that have been unchecked. I think people really want to have a conversation about, “Is this what we want? We didn’t vote for this.”
You trace this history back long before Silicon Valley. Why?
Lepore: We don’t really have much of a sense of how we got here, and it’s very tempting to think we started getting here in 1995. It’s also tempting to engage in conspiratorial thinking, like, “Oh, the state has been hijacked by these corporations,” when in fact so many elements of the infrastructure of what I call the artificial state were designed with the best of intentions by good people just trying to help democratically elected government be more efficient.
Counting the population for purposes of proportional representation more accurately, keeping track of Veterans Administration records to better provide veterans’ benefits, or using computers to administer Social Security and taxes are not malign or sinister developments.
What I try to pay attention to is: Is this thing actually replacing something that’s really important to how people belong to a society? Public-opinion polling would be a good example. Over time, doing the “press a button” polling replaces a lot of those face-to-face interactions. Think about all those hundreds, thousands, millions of conversations: chatting at the side of the football field with the parents of the team. That’s just chit-chat. That’s part of how you belong in a community.
Where do you draw the line between useful automation and surrendering human judgment?
Lepore: When I play around with Claude or something, and I think about the incredible applications of this for oceanography or understanding animal communication or gestural languages of elephants or public-health research, I’m just really dazzled. I wish we could have a lot more excitement that ordinary people could share in that excitement: “Wow, this is really great. Look at this incredibly cool tool that can do all these things.”
But the pernicious stuff is more like using large corpora of data to predict social and economic outcomes for people. Remember when ProPublica did this huge piece on black-box sentencing algorithms that a lot of courts were using? There was such a strong reaction to that reporting partly because it was proprietary software, someone was making money off doing this, and no one could assess whether it was just or right.
Our criminal courts are really backed up. This made things faster. But their reporting suggested that it reproduced biases that were already in the criminal-justice system and therefore perpetuated and amplified those biases.
If this kind of intelligence and predictive analytics is going to be used to determine the fates of individuals, those are some of the boundaries that I think, if we were slowing things down and not having our fates determined by these corporations and their agendas, we would have been making better decisions about.
So is the central problem the technology, or the concentration of power around it?
Lepore: I think people are not making a ton of choices right now. I think that’s part of what the growing backlash is about: people feeling like they didn’t choose this, but it’s somehow inescapable.
At work, “Oh, you must use this now.” At school, “Well, your kids are going to have this now.” Or now your town’s going to have this data center. Which is a long way of saying I think it’s about the power, the unchecked power of people who are interested in advancing their own fortunes over any accountability to a democratic process.
Interview by Sascha Brodsky. This conversation was originally conducted for reporting in 2026 and is published here in an edited Q&A format.
AI Safety Watch 