The Case for Letting the Public Argue With AI, Not Just Trust Its Builders

Nathan Gardels, writing in Noema, argues that decisions about how AI systems are built and deployed are being made almost entirely by a handful of technology companies and their algorithmic agents, with no equivalent yet of the public deliberation that eventually tamed earlier communication revolutions. He traces a pattern: the printing press, mass newspapers, and broadcast media all caused real social disruption before civil society, regulators and market competition brought them into a rough consensus. Gardels warns that AI, and especially AI-curated social media, is moving far faster than those precedents, leaving little time for the same self-correcting process to unfold on its own. His proposed remedy is public feedback at genuine scale — not symbolic consultation, but structures that let ordinary users contest and shape how these systems actually behave. He points to the Reformation, the Enlightenment and the early mass-media era as reminders that such adjustments took decades and real turmoil before they stabilized. The argument lands one day after Anthropic's own admission that it is withholding an internal model it judges too risky to release — quiet proof, on Gardels's account, that even AI's builders don't yet trust their own governance of it.