The Internet Revolution We Misread
We predicted the internet would bring democracy, abundance, and connection. It brought all three and also their opposites. Understanding what we got wrong helps explain what comes next.
John Perry Barlow's "Declaration of the Independence of Cyberspace," written in 1996 and addressed to the governments of the industrial world, announced: "Cyberspace does not lie within your borders... You have no sovereignty where we gather." The declaration was simultaneously prescient and naive — prescient in recognizing that digital networks were creating governance challenges that territorial states were unprepared for, naive in assuming that the absence of government sovereignty meant freedom rather than the replacement of state power with corporate power.
The internet optimism of the 1990s and early 2000s — the technolibertarian conviction that digital networks were inherently democratizing, that they would lower barriers to communication and organization, that they would undermine centralized control and enable horizontal, distributed political and economic organization — was not simply wrong. It described real things that were really happening. But it systematically underestimated two forces that were also really happening: the tendency of digital networks to concentrate power in platforms, and the malleability of digital tools for purposes other than liberation.
What the optimists got right
The democratic potential that the early internet promised materialized, imperfectly, in several ways. The Arab Spring protests of 2011 demonstrated that social media platforms could coordinate collective action against authoritarian governments at speeds that traditional organizing could not match. Wikipedia, for all its failures, created a genuinely unprecedented corpus of knowledge accessible to anyone with internet access. Open source software development organized by distributed contributors produced operating systems, programming languages, and applications that undermined the monopoly models that had dominated the technology industry.
The democratization of content creation that the internet enabled was also real: the barriers to publishing text, audio, and video that had maintained the gatekeeping functions of traditional media essentially collapsed. The long-tail hypothesis — Chris Anderson's 2004 prediction that digital distribution would allow niche markets to sustain themselves and reduce the dominance of mass-market products — was partially correct.
The counterfactual question — would political freedom have been worse in the absence of the internet? — is impossible to answer definitively but is probably yes in many countries. The internet has provided organizing tools that authoritarian governments have found easier to use than to simply suppress.
What the optimists got wrong
The optimists underestimated the platform concentration dynamic. The internet's network effects — the tendency of platforms to become more valuable as they grow, creating winner-take-most competitive dynamics — were legible from economic theory well before their empirical realization, but the optimists consistently predicted that network effects would dissipate rather than reinforce.
Instead, Google captured search and online advertising; Facebook captured social graphs; Amazon captured e-commerce and cloud computing. The same network dynamics that had made the internet a distributed challenge to existing power produced a new set of concentrated powers that are in many respects more difficult to challenge than the governments the optimists had been worried about.
The optimists also underestimated the malleability of digital tools for authoritarian purposes. The social media tools that enabled the Arab Spring also enabled the counter-mobilization that suppressed it; the same networks that distributed dissident information distributed state disinformation; the same organizing platforms that facilitated democratic coordination facilitated authoritarian coordination. Russia's Internet Research Agency and China's 50 Cent Army demonstrated that digital networks could be weaponized against democracy as effectively as they could be used for it.
The AI moment as a replay
The discourse surrounding artificial intelligence in 2024-25 bears significant structural resemblances to the internet discourse of 1995-98: the confident predictions of democratization and liberation, the insufficient attention to concentration and control dynamics, the underestimation of the technology's malleability for purposes other than its proponents intend.
The AI optimists who expect that large language models and autonomous agents will democratize access to legal advice, medical expertise, and educational support are probably right that these applications will develop and that they will provide real benefits to underserved populations. The concentration dynamics that are already visible — three or four US companies controlling the most capable foundation models, the enormous capital requirements for training and inference, the regulatory moats that established players will work to construct — are probably also going to be real and possibly dominant.
The question of whether AI's net effect on democracy, inequality, and human autonomy will be positive or negative depends heavily on governance choices that are being made now, in a period of enthusiastic optimism that mirrors the early internet period in its insufficient attention to the things that can go wrong. The internet's history is not a prediction of AI's future; it is a set of precedents that the AI moment is not taking seriously enough.
Cara Novak is a contributing writer at The Auguro covering European history, democratic theory, and comparative politics.