How is AI shaping democracy?

What It Covers

Bruce Schneier, fellow at Harvard's Berkman Klein Center, discusses his book "Rewiring Democracy" with hosts Daniel Whitenack and Chris Benson. The conversation examines how AI transforms elections, legislation, government administration, courts, and citizen engagement across global democracies, moving beyond deepfakes to explore substantive democratic applications.

Key Questions Answered

• •Public AI Models: Switzerland's ETH Zurich created a competitive core model using government funding and supercomputer resources without corporate profit motives, illegally stolen training data, or exploited third-world labor for fine tuning. This demonstrates viable alternatives to hundred-million-dollar corporate models, with expectations of twenty similar public models emerging within a year.
• •AI-Enhanced Political Engagement: Japan's Takahiro Anno built an AI avatar to answer voter questions on YouTube during his Tokyo mayoral campaign, finishing fifth among fifty candidates. He later won election to Japan's upper legislative chamber and now uses AI through his Team Mirai party to facilitate constituent discussions on legislation and priorities across Japanese parliament.
• •Judicial Efficiency Through AI: Brazil implemented AI systems to manage court operations, assigning judges to cases and moving documents through the system. This reduced judgment times from two to three years down to faster resolutions, though attorneys simultaneously use AI to file more cases, ultimately increasing access to justice despite higher case volumes.
• •AI-Assisted Investigative Journalism: California's CalMatters collects every public statement from elected officials including floor speeches, campaign emails, and tweets, then uses AI to identify anomalies in voting records and campaign contributions. The system creates tip sheets exclusively for human journalists to research and verify stories, not publishing AI findings directly.
• •Decentralization Through Cost Reduction: DeepSeek demonstrated competitive AI models require neither cutting-edge chips nor massive funding, challenging the assumption that only five US monopolies can develop effective AI. Specialized models for specific tasks like physics teaching or restaurant recommendations will replace massive general-purpose models, with personal butler models coordinating access to specialized systems.