The prospect of artificial intelligence destroying humanity has become a prominent part of the technology industry’s public debate, but Harvard historian Jill Lepore argues that the more immediate danger lies elsewhere: the gradual transfer of human decisions and relationships to machines. In an interview with German newspaper Die Zeit, Lepore said governments and technology companies should take AI seriously without allowing predictions of catastrophe to replace sober political debate.
Lepore, whose new book Regierung der Maschinen. Aufstieg und Fall des künstlichen Staates examines the effects of artificial intelligence on government and society, said the current discussion reminded her of earlier episodes of public panic. She recalled being asked by The New Yorker in 2009 to write about swine flu amid warnings of a potentially catastrophic global pandemic. Although those predictions later proved substantially exaggerated, she remembered walking across Harvard’s campus wondering whether she was about to become a victim of a pandemic. For Lepore, such experiences illustrate what she calls humanity’s persistent appetite for stories of catastrophe.
She does not dismiss AI as a risk. “There is a risk, certainly also from AI,” she told Die Zeit, while stressing that even leading AI experts cannot reliably assess the danger. She compared the current debate with what she calls the “epidemic pitch”, a strategy used by US scientists in 1929 to warn the public about psittacosis and persuade the government to establish institutions for risk research and preparedness. That intervention, she said, ultimately served the public interest. From a historian’s perspective, she argues that the leaders of AI companies have also wanted the state to intervene, while warning that such companies should not rely on apocalyptic narratives to achieve that goal.
Her own concern is the transformation of everyday life and political institutions. Silicon Valley companies have repeatedly promised that technological innovation would make people happier, healthier and less poor, while helping create freer and more democratic societies. Lepore believes many people came to regard that promise as broken, particularly after the rise of social media. She also points to the enormous AI data centres now being constructed. “I myself constantly hold this little, insanely helpful thing in my hands, and at the same time I have the feeling that I am chained to a monstrous machine from which I cannot get free,” she said. “It is as if I am losing control over my life.”
Lepore describes the emerging possibility of an “Artificial State” in which AI systems and AI agents increasingly make decisions for and about citizens. Liberal democracy, she argues, depends on self-determination: people elect governments and participate in deciding how they live together. The introduction of automated decision-making therefore has implications beyond technological efficiency.
She sees evidence of this shift particularly in the public sphere. An increasing share of political debate takes place on online platforms, where algorithms and increasingly AI systems influence who becomes visible and who receives attention. Lepore said there are already more bots than humans on some platforms and argued that treating platform users as a perfect representation of the electorate can distort political understanding. Direct encounters between politicians and voters have declined, she said, while platform-based discussion tends towards polarisation.
The same issue, in her view, extends to local politics. Online participation can make public meetings more accessible, but physical meetings also bring people with opposing views into direct contact, creating opportunities to resolve disagreements before they escalate. “Platforms turn citizens into angry trolls,” she said, describing this as a loss for civil society.
Lepore also disputes the assumption that greater automation necessarily means better public services. She noted that during a recent arrival in Italy she was processed by a machine rather than a person. More broadly, she warned that personal and analogue experiences could increasingly become privileges of wealthier people, while poorer citizens receive automated services. In the United States, she said, some patients can no longer obtain personal medical appointments and instead interact with AI systems. Income could therefore help determine whether a person receives time with another human being or an AI diagnosis, including automated interpretation of laboratory results.
Even if AI systems were assumed never to make mistakes in selecting people or diagnosing patients, Lepore said, the social consequences of removing human interaction from more areas of life would remain significant. She made the same argument about education. If Harvard students perform better with an AI chatbot than with a human tutor, she said, that does not mean the tutor’s role has become unnecessary. Conversations with tutors teach communication with people who possess greater knowledge, while tutors also introduce students to the wider university community. Replacing those relationships with AI, she said, would amount to “an intellectual capitulation”.
Lepore draws on Hannah Arendt’s analysis of totalitarianism, particularly Arendt’s concerns about automation, mass media and the weakening of traditional social structures. She does not claim that AI will necessarily destroy liberal democracy, saying instead that she hopes warnings about the technology will stimulate debate and strengthen society’s capacity to resist harmful developments.
She is also sceptical of the technological optimism associated with parts of Silicon Valley. Lepore criticised Anthropic founder Dario Amodei’s use of a Richard Brautigan poem imagining a harmonious future in which mammals and computers coexist. She argued that the poem was intended as satire of the Californian ideology of the 1960s rather than as a technological utopia. She also criticised the energy consumption of AI companies, arguing that data centres powered by coal and gas contribute to the climate crisis.
For Lepore, the response is not to reject technology but to establish rules governing its use while recognising individual responsibilities within democratic society. “That means, for example: don’t post, but go there. Get off the sofa,” she told Die Zeit.
