Liberal societies have celebrated their openness. Public records, hearings, and archives are available to anyone who wishes to look, but few can access them. Information may be legally public, but practically unavailable due to the high costs of accessing and analyzing that information. The high cost deterred transparency as interlocutors could, in principle, read everything, but not efficiently. Records were open, but few could extract the signal from the noise.
Democratization of Knowledge
Large language models (LLMs) offset that old imbalance by dramatically lowering the cost of search, interpretation, and cross-referencing. In short, LLMs turn openness into accessibility. Following Ronald Coase’s insight that information and coordination are costly, firms exist when it is cheaper to organize internally than to transact through the market. The same principle applies to information environments: LLMs reduce the cost of searching, translating, acquiring, and otherwise transforming text. Keyword search and reading scale linearly with volume, but LLMs shift the slope by vastly reducing the marginal cost of reviewing another million words. The practical effect is that what was previously hidden by abundance becomes accessible.
The most important shift is not technological, but conceptual: LLMs will turn accessibility into usability across domains like governance, business, and science. To illustrate the point:
Public oversight: The public sector generates enormous quantities of text, much of which remains effectively unread. A suitable machine prompt can reveal recurring justifications for sole-source contracts, identify departments that repeatedly miss reporting deadlines, or trace changes in budget priorities over time. The shift is from anecdotal oversight with citizens and journalists focusing on isolated cases, to oversight where statistical anomalies become visible at once. The costs of public scrutiny fall dramatically.
Legal-regulatory: Administrative law is a paradigmatic case of information overload. Rulemakings, consent decrees, settlement agreements, and comment dockets are all public but scattered across incompatible systems. LLMs can flag upcoming sunset clauses or overlapping mandates that create regulatory conflict. For legal practitioners, this automation reduces the cost of compliance; for scholars and watchdogs, it exposes patterns of regulatory drift or capture that were previously almost invisible.
Scientific and policy translation: Scientific and policy domains generate vast textual ecosystems — preprints, guidance documents, technical reports — filled with overlapping claims. LLMs can rapidly extract key propositions, map networks of agreement and dissent, and link findings to underlying evidence. Instead of reading thousands of pages, policymakers and researchers can query a summary of consensus and controversy, reducing the cost of accessing expert discourse without eliminating the need for it.
Friedrich Hayek observed that knowledge in society is frequently dispersed and tacit. Prices coordinate diverse sets of knowledge by summarizing information that no individual fully possesses. LLMs perform a similar coordination task for text. They compress the distributed linguistic record of human activity into digestible, context-specific summaries (of greater or lesser levels of objectivity). The wealthy have always been able to afford to hire lawyers, researchers, and consultants, but those without such means faced much higher coordination costs. LLMs make “good-enough” expertise available to anyone with a smartphone, lowering the effective cost of participation in markets and civic life. This structural shift in access to knowledge makes information more available to those who are less well-off.
How Accessibility Redefines Power
Unfortunately, though, these same properties that empower individuals can weaken collective safeguards. Liberal democracies have long relied on informational clutter as a de facto security feature, but with the rise of LLMs, adversaries are better able to exploit public data. They always could, but now they can do it easily and orders of magnitude faster. Automated summarization and cross-linking make the full corpus of open information vastly more accessible to anyone with access to LLMs. What was once an advantage — transparency coupled with scale — can become a liability. Accessibility and transparency cut both ways.
Coase’s framework predicts institutional reorganization when transaction costs fall, and we are already seeing that. Apple’s decision to outsource generative AI instead of engineering its own reflects a classic make-versus-buy calculation to reduce internal coordination costs. Similar adjustments will occur in universities, law firms, and agencies. The appropriate response, then, is adaptation to those reduced costs.
Safeguards for an Accessible Society
As such, liberal societies can no longer rely on the sheer abundance of data to protect against misuse or distortion, because LLMs undercut that defense. Valuable information that was either buried in plain sight or incredibly costly to find and acquire is more visible and accessible. This transformation will expose previously hidden inefficiencies and opportunities for accountability, but it will also expose strategic vulnerabilities. Our mission must be to ensure that gains in knowledge accessibility do not come at the expense of political stability or individual privacy.
LLMs represent a turning point in the political economy of information by reducing the costs of finding, understanding, and connecting the text in the same way the printing press reduced the cost of reproducing it. The result is more usable information. Whether this technological shift will strengthen or weaken liberal societies depends on institutional adaptation. If liberal societies remain too open, without taking appropriate steps, the advantages of accessibility will be offset by new forms of exploitation. If, however, liberal institutions evolve to embed verification and resilience, LLMs will extend rather than erode the project of Enlightenment liberalism.