Rebooting Democracy in the Age of AI

This week on Reboot Democracy, Elana Banin examines World Health Organization research on how AI could transform evidence-informed policymaking, while warning that the harder challenge is whether ministries can build the capacity to align with frontline realities. Amedeo Bettauer argues that the public conversation on AI is being shaped by a narrow “messenger class,” leaving students, workers, and families already using these tools largely absent from the debate. And the Center for AI and Digital Policy's AI Index finds that, across 90 countries, governments increasingly legislate AI governance principles such as fairness and transparency; the gaps lie in implementation, oversight, and public-sector capacity. Elsewhere, states experiment with AI in benefits systems and child welfare, the Labor Department prepares a workforce data hub to track AI’s economic effects, and South Africa withdraws its national AI policy after fabricated AI-generated citations expose the risks of weak institutional review. This and more in this week’s News That Caught Our Eye.

The 2026 CAIDP AI Index, ranking AI policy commitments across 90 countries, shows that while most governments now agree on core governance principles, the real divide lies in implementation. Many are advancing laws, oversight, and public participation, but progress lags in turning commitments into practice. As the baseline shifts from whether to govern AI to how, the report underscores that outcomes depend less on frameworks and more on the capacity of public institutions—and the civil servants within them—to operationalize these principles in everyday decisions.

A new WHO discussion paper explores how AI can accelerate research synthesis and keep evidence continuously up to date. Elana Banin welcomes the push to use AI to strengthen the evidence-to-policy pipeline, but argues the more consequential question is whether AI will redefine what counts as evidence in the first place. The harder constraint will ultimately be institutional, as most health workers lack the training and infrastructure to adopt these tools. Government decision-makers must start building the processes to test AI outputs against frontline knowledge and the capacity to make that adoption defensible.