Jill Lepore’s “artificial state” warning—and why it lands now
Harvard historian Jill Lepore, speaking recently on the “Hard Fork” podcast, frames today’s AI moment as more than a technology cycle. Her central claim is a political one: artificial intelligence is accelerating the emergence of an “artificial state,” a system in which corporate-owned machines and algorithmic infrastructures increasingly shape public life—from what people see and believe to how decisions get made—while democratic institutions struggle to keep pace.
The provocation is not that governments will vanish, but that sovereignty subtly migrates. When citizens rely on AI intermediaries for information, advice, and even quasi-civic functions, the practical authority to set agendas and define “reasonable” outcomes can drift toward the entities that own the models, the data pipelines, and the distribution channels. Lepore’s critique directly challenges a familiar Silicon Valley storyline: that AI’s spread is inevitable, neutral, and inherently democratizing. Her historical lens argues the opposite is at least as plausible—especially given how the internet and social media once carried similar promises, only to consolidate into platform power, surveillance advertising, and polarization.
At the heart of Lepore’s argument is a cultural and institutional concern: public life becomes “factory farmed”—optimized, standardized, and scaled—while the messy, local, participatory texture of civic engagement thins out. The risk is not only misinformation or bias in outputs, but a deeper reconfiguration of how legitimacy is produced: less through deliberation and accountable institutions, more through automated systems that nudge, rank, recommend, and decide.
Algorithmic sovereignty: when platforms start behaving like institutions
The most consequential shift Lepore points to is the quiet normalization of algorithmic sovereignty—the idea that AI systems increasingly perform roles once reserved for elected bodies, courts, professional guilds, or community institutions. This is not limited to high-stakes “AI in government.” It also appears in everyday, cumulative ways: recommendation engines shaping political attention; generative AI shaping how people interpret policy; automated moderation shaping the boundaries of acceptable speech.
Several dynamics make this transition structurally plausible:
- AI-mediated discourse at scale: Social platforms and AI assistants increasingly mediate news, debate, and identity formation. The “public square” becomes a set of privately governed interfaces.
- Personalized information diets: When civic information is filtered through behavioral profiles, shared reference points erode—weakening the common ground that democratic deliberation requires.
- Delegation by convenience: People outsource tasks to AI because it is fast and frictionless. Over time, convenience becomes a governance mechanism: what is easiest becomes what is normal, and what is normal becomes what is legitimate.
Lepore’s warning is sharpened by precedent. The early internet was widely associated with decentralization and empowerment; social media was pitched as democratizing voice and participation. Yet the dominant outcome was consolidation—a small number of platforms became chokepoints for attention, commerce, and communication. The AI era, in her telling, risks repeating that arc, but with higher stakes: not only distributing speech, but synthesizing knowledge and recommending action.
The infrastructure story: hyperscale data centers as the new civic backbone
A defining feature of the “artificial state” is not just software, but infrastructure consolidation. The rapid buildout of hyperscale data centers and cloud services centralizes compute and data in the hands of a few global firms. These facilities increasingly function like critical national infrastructure, yet remain largely governed by corporate incentives and cross-border supply chains.
This concentration matters for business, government, and society because it creates new gatekeepers:
- Compute as a strategic bottleneck: Access to advanced chips, cloud capacity, and model training pipelines becomes a competitive and geopolitical constraint.
- Data as a quasi-public resource—privately held: Behavioral data, communication metadata, and content flows become the raw material for influence and monetization.
- Service layering over public functions: As AI vendors bundle messaging, identity, authentication, and decision support, they can become embedded in workflows that resemble public administration—without the same transparency or accountability.
Lepore’s historical analogy to the consolidation of utilities in the late 19th century is instructive for executives and regulators alike. Utilities eventually triggered public oversight, antitrust action, and rate regulation because society recognized them as essential services. AI infrastructure may be approaching a similar threshold—except its “rates” are paid in attention, autonomy, and governance capacity, not only dollars.
Markets, labor, and geopolitics: the business incentives behind the drift
The “artificial state” is not merely an ideological project; it is also a market outcome. Capital markets and corporate strategy currently reward scale, data capture, and platform lock-in. That incentive structure can pull AI deeper into civic and institutional domains, even when the stated mission is consumer convenience or productivity.
Key implications emerge across stakeholders:
- Corporate capture of quasi-public functions: AI platforms can expand into dispute resolution, hiring, education, benefits navigation, and even voting-adjacent services—turning civic tasks into product surfaces.
- Regulatory arbitrage: Companies can test and deploy systems in jurisdictions with weaker AI governance, creating uneven standards and a “race to the bottom” dynamic.
- Investment concentration: Venture capital and corporate R&D increasingly cluster around AI, inflating valuations tied to network effects and long-run data monetization assumptions.
The labor market dimension is equally central. Knowledge work automation does not simply eliminate tasks; it can reorder professional hierarchies. Organizations that master AI workflows gain leverage, while labor markets risk bifurcating between highly specialized technical roles and a broader service class supporting AI-driven enterprises. Lepore’s “factory farming” metaphor resonates here: optimization at scale can reduce human agency, even when productivity rises.
Geopolitically, the trajectory points toward competing visions of digital sovereignty: export controls, data localization, domestic AI champions, and standards battles. The question is no longer whether AI will be regulated, but who sets the rules—and whose values are encoded in the defaults.
What Lepore ultimately surfaces is a governance choice disguised as a technology trend. If AI becomes the interface through which citizens understand the world and institutions deliver services, then the architecture of AI becomes the architecture of democracy—or its substitute. The next phase will be defined less by model benchmarks than by whether societies can build accountability, interoperability, and human-centered checks before the “artificial state” hardens into the background operating system of public life.
By
By
By
By
By
