Insights on AI governance, public sector innovation, and democracy
AI and Public Engagement
Public input builds trust when someone has the authority, and the duty, to act on it.
California's AI-EP project made parents co-designers of an AI tool that makes their children's education plans easier to understand. More than 200 families joined the first pilot and came away better able to advocate for their kids. It shows that who you choose to include shapes who the tool serves.
California's AI-EP project made parents co-designers of an AI tool that makes their children's education plans easier to understand. More than 200 families joined the first pilot and came away better able to advocate for their kids. It shows that who you choose to include shapes who the tool serves.
The Hugging Face incident wasn't a rogue AI escaping its sandbox. It was a chain of poor human decisions. We need to stop talking about AI like mythology and start treating it like engineering.
The Hugging Face incident wasn't a rogue AI escaping its sandbox. It was a chain of poor human decisions. We need to stop talking about AI like mythology and start treating it like engineering.
Governors from Healey to Shapiro now say data centers need local buy-in, but they haven't said what a community "yes" actually looks like. Since 2024, at least 46 projects representing more than $170 billion have been blocked, withdrawn, or stalled, often because engagement started after the key decisions were made and benefits were whatever developers chose to offer. Deals in West Virginia, Louisiana, Texas, and Lancaster, Pennsylvania show the terms can be negotiated, but only if states give communities the capacity to take part: early information, independent expertise, and a process that developers can pay for without owning.
Governors from Healey to Shapiro now say data centers need local buy-in, but they haven't said what a community "yes" actually looks like. Since 2024, at least 46 projects representing more than $170 billion have been blocked, withdrawn, or stalled, often because engagement started after the key decisions were made and benefits were whatever developers chose to offer. Deals in West Virginia, Louisiana, Texas, and Lancaster, Pennsylvania show the terms can be negotiated, but only if states give communities the capacity to take part: early information, independent expertise, and a process that developers can pay for without owning.
A weekly roundup of the most intriguing and impactful stories from the world of artificial intelligence
After the 2025 Los Angeles wildfires, California's Office of Data and Innovation used Engaged California to gather more than 1,300 resident comments. AI helped turn that input into 19 policy options, which residents then deliberated over and ranked. Jeffery Marino explains how clear ownership, with state agencies, local governments, and community groups acting together with support from the Governor's office, turned those priorities into a five-part recovery action plan.
California's AI-EP project made parents co-designers of an AI tool that makes their children's education plans easier to understand. More than 200 families joined the first pilot and came away better able to advocate for their kids. It shows that who you choose to include shapes who the tool serves.
The Hugging Face incident wasn't a rogue AI escaping its sandbox. It was a chain of poor human decisions. We need to stop talking about AI like mythology and start treating it like engineering.
Governors from Healey to Shapiro now say data centers need local buy-in, but they haven't said what a community "yes" actually looks like. Since 2024, at least 46 projects representing more than $170 billion have been blocked, withdrawn, or stalled, often because engagement started after the key decisions were made and benefits were whatever developers chose to offer. Deals in West Virginia, Louisiana, Texas, and Lancaster, Pennsylvania show the terms can be negotiated, but only if states give communities the capacity to take part: early information, independent expertise, and a process that developers can pay for without owning.
At least 46 AI data center projects — more than $170 billion in announced investment — have been blocked, withdrawn, or stalled since 2024 following community opposition. In this week's Reboot piece, Beth Simone Noveck and Deborah Stine argue trust is the infrastructure data centers are missing, and developers can pay for the community conversation without owning it. Rutgers' National AI Opinion Monitor shows why: AI adoption hit 72% in August while worry climbed to 50% and anger to 30%. NYC's Public Advocate has called for an AI bill of rights; the Maldives' Prosecutor General now requires courtroom disclosure of AI use; and The American Prospect reports Anthropic has been running a predictive surveillance system that flagged suspected dissenters to police.
Ben Peterson, Executive Director of the City-County Planning Commission of Warren County, Kentucky, on how What Could BG Be? used AI-supported tools to turn input from nearly 8,000 residents into Bowling Green's next 25-year comprehensive plan.
Three-Year Multistate Initiative Will Combine Free Mobile AI Learning with Library-Based Programs Helping Residents Navigate Government, Participate in Civic Life, and Solve Community Problems with AI
As this week's curator of News That Caught Our Eye, I'm sharing the latest resources available to help public officials think through a proposed data center. Every AI data center and every community is different. These resources provide the tools to help each community think through its decisions about an AI data center, often in a challenging environment.
A new free course from the Allen Lab for Democracy Renovation, The GovLab, and InnovateUS teaches public servants how to design AI-enabled public engagement that turns comments into policy, drawing on 42 real-world examples from 20 countries.
The public conversation about AI has swung from exuberance to alarm, casting AI as an unstoppable force. That framing overlooks the fact that public institutions design, build, buy, and regulate AI, and societies still choose where it leads. This fall, Reboot is organizing its coverage around four connected questions: what capabilities public institutions need, how they can become institutions that listen and learn, what public AI infrastructure requires, and how we measure whether AI is actually creating public value.
AI is making it possible for governments to listen at a scale that was previously impossible, turning thousands of comments, service interactions and other signals into usable knowledge. But automated synthesis is not neutral: every act of compression determines what becomes visible, what disappears and ultimately what institutions come to know. If AI is to strengthen democracy rather than flatten it, AI-assisted listening must meet a higher standard: fidelity to what people actually said, preservation of plurality and disagreement, traceability back to the underlying evidence, contestability by those being represented, and a clear connection between what is heard and consequential decisions.
Two trends are unfolding in parallel around the world. Public anxiety about artificial intelligence is growing while communities’ satisfaction in how they participate in civic activities and contribute to government policy and programs is declining. Pathways to resolve this tension are emerging as governments find ways to use these technologies to help communities participate more meaningfully. Examples from Bogotá, Hamburg, Brazil, Mexico and elsewhere show how AI can improve the quality and experience of public engagement while reducing its time and cost. If done poorly, or without safeguards, it can also introduce new risks. However, these can be managed with appropriate process design, regulation and delivery by skilled practitioners.
A Boston Globe editorial argues that Massachusetts should use AI to identify outdated laws, unnecessary reporting requirements, and inefficient regulations, drawing on reforms and pilot programs in other states. Featuring insights from Beth Simone Noveck and Stanford RegLab, it explains how AI-assisted review, paired with human judgment, could reduce administrative burdens and improve government efficiency.
As prediction increasingly shapes decisions about jobs, healthcare, education, and public services, the next frontier is how it organizes society. Drawing on the history of the railroads, this essay argues that policymakers already have a blueprint for governing essential technologies: treat AI as infrastructure, apply public oversight, and ensure that systems central to modern life serve the public interest rather than private power.
At 92NY’s State of America Summit, Beth Simone Noveck, Betsey Stevenson, and Allison Pugh discuss AI, the future of American labor, and whether technology will strengthen public voice or erode human connection.
InnovateUS 