Insights on AI, Governance and Democracy
What happens when governments rely on systems that sound right instead of being right? Drawing on OpenFisca’s spread from France to governments across Europe, Africa, and Oceania, Beth Simone Noveck’s interview with Matti Schneider makes the case for public infrastructure that computes the law, as well as the risks of sidelining it as generative AI scales globally.
Using AI to analyze over 2,400 employee comments, California’s Engaged California team found that the challenge wasn’t the scale of the data, but making sense of complex, layered input without oversimplifying it. Their experience shows why human judgment remains essential, from building taxonomies to catching errors, as results can shift significantly depending on how AI is applied and what people choose to trust and prioritize.
AI governance is expanding fast, but not evenly. A new analysis from MIT and Georgetown’s CSET maps over 1,000 governance documents to show that while policies are proliferating, they cluster around familiar risks and sectors, leaving key gaps across socioeconomic impacts, upstream design, and everyday domains. The result, as relayed by research member Yan Zhu, is a more precise picture of what AI governance actually covers, what it still overlooks, and where policymakers should focus in the future.
Across contexts like India, where author Anirudh Dinesh’s family lives, AI chatbots such as xAI’s Grok are increasingly used not to inform but to generate arguments that reinforce existing political views, creating “generative echo chambers.” Unlike passive social media exposure, users actively prompt AI to validate positions, often producing confident but inaccurate claims that go unchecked. While some research suggests AI can moderate views in neutral dialogue, real-world use skews toward advocacy, compounded by low verification and high trust in outputs. The result is that AI may not just reflect polarization, but actively deepen it, depending on how these systems are designed and used.
More than $1 trillion in federal grants flows to communities each year, but complexity keeps much of it out of reach. This piece by Beth Simone Noveck, published by Fast Company, explores how AI can either deepen that gap or help close it. The solution is GrantWell, a community-centered tool designed with local governments to make funding accessible and public systems work as intended. Launched in Massachusetts and expanding to additional states, it shows how AI can help communities claim the resources already set aside for them.
In “Building a Human Resilience Infrastructure for the AI Age,” based on insights from 386 global experts, a warning emerges: the greatest risk is not a single catastrophic event, but a slow drift toward diminished human agency, fragmented reality, and growing dependence on automated systems. Lee Raine and Janna Anderson reflect on how, as AI becomes society’s invisible operating system, individual resilience is no longer enough. The report argues for urgent, coordinated action to build resilience as shared infrastructure across governments, institutions, and communities to ensure people can still question, contest, and shape the systems increasingly shaping them.
Developed by the International Training Centre of the International Labour Organization with and for employer organizations, ReguLens responds to a growing policy bottleneck. The rapid surge in complex, overlapping regulations is outpacing institutional capacity. Built through iterative co-creation with users across regions, the tool helps organizations analyze proposals, identify impacts, and engage earlier in policy debates.
Stefaan Verhulst and Adam Zable argue that the biggest challenge in open data is no longer access, but usability, as official statistics remain difficult to find, interpret, and apply. Drawing on the International Monetary Fund’s StatGPT new research, it shows how artificial intelligence could transform access through natural language interfaces, while warning that accuracy and trust depend on retrieving authoritative data rather than generating answers. The article situates this shift within a broader “Fourth Wave of Open Data,” calling for new data systems and governance approaches that make information truly usable and reliable.
Following the success of a 1,200-person pilot across eight state agencies, New York is scaling InnovateUS' AI training initiative to more than 100,000 state employees, the largest program of its kind in the nation. Read the NYS press release detailing the expansion, which focuses on helping public servants build the skills and knowledge to responsibly use emerging technologies for the public good.
This piece by John Wihbey and Jill Abramson distills lessons from the InnovateUS Amplify workshop series, where public-sector communicators grappled with how AI is reshaping their work amid declining public trust. Across sessions, we learn that AI can accelerate research, synthesis, and storytelling, but it cannot replace judgment, verification, or institutional values. As communicators adopt new tools, the opportunity lies in using AI to strengthen transparency, credibility, and connection with the public, rather than erode them.
Designing a one-hour course on democratic engagement and AI means confronting genuinely hard questions about representativeness, political framing, and audience, where thoughtful experts disagree, and every choice involves a real tradeoff. Over the past week, we drafted, debated, and cut more than 25,000 words to a working script, informed by over 300 comments from 50 advisors across 24 countries and a room full of democratic theorists in Barcelona. This post explains the three hardest calls we had to make and why we made them.
Millions of Americans miss out on health and food assistance benefits due to fragmented systems and complex enrollment processes. This piece explores how Link Health, in partnership with the AI for Impact program, is combining AI tools with human navigators to rethink how public benefits are delivered in healthcare settings. It argues that the next frontier is better evidence. States should fund research to compare enrollment approaches, portal design, and navigator support to determine which improve health outcomes and guide smarter public investment.
AI agents are reshaping how systems operate across sectors. This piece argues that the imminent challenge to address is autonomy, including how agents act, interact, and scale in open environments. It outlines a three-part blueprint for governments to build trust infrastructure, prepare for multi-agent risks, and develop the institutional capacity needed to govern an increasingly agentic world.
Growing up in Kakuma refugee camp, Nhial Deng experienced what it means to be excluded from opportunities. Returning years later, he saw young people using AI not as aid, but as a tool to build skills, income, and futures in real time. This piece argues that AI is already functioning as an economic opportunity layer, but one that remains uneven and fragile without intentional design. Drawing on examples from Canada, Singapore, and Kenya, Deng outlines how governments can move from accidental access to structured opportunity by connecting AI to jobs, embedding it in trusted institutions, and building safeguards alongside deployment.
The Healey-Driscoll Administration has launched GrantWell, a free AI-powered tool developed with Northeastern University’s Burnes Center for Social Change to help municipalities more easily access and apply for federal and state funding. GrantWell helps you summarize complex grant requirements, identify opportunities, and draft early-stage proposals, reducing administrative burden and expanding capacity to secure resources
InnovateUS