Kris Rose is a Georgetown McCourt School of Public Policy Tech & Policy Visiting Fellow. He is an AI transformation leader at IBM and previously held governance and policy roles at other technology companies and the US government. The views expressed here are his own and do not reflect those of his employer.

Protests. Boos at commencements. Warnings from the pope. It’s hard to miss Americans’ increasingly palpable anxiety about AI, even as it becomes more and more a part of daily life.

Why is this the case? I think the paradox is clear: we are building the future, but we are doing it without the people we are building it for.

The reality is that there are two crises of confidence surrounding artificial intelligence. The first, as noted above, is the profound public anxiety about the technology itself, which has intensified since ChatGPT's release in 2022. Hardly a day passes without a new poll pointing to the public’s concern about AI’s impact on things like job displacement, mental health, misinformation, and environmental impact. Axios has termed it an “AI hate wave.”

The second crisis is one of legitimacy facing the institutions that are responsible for developing and governing the technology. Americans' trust and confidence in the federal government and big tech executives have cratered. While nearly 8 in 10 Americans think it’s urgent for the government to pursue AI regulation, there is little consensus on how and who can be trusted to do so. One survey found that Americans trust independent experts over the federal government and industry to address AI guardrails – respondents identifying as Democratic or Independent were significantly less trusting of the US federal government’s ability to do so, likely because of Republican majorities in DC. Taken together, Americans urgently want action, but there is little trust that our existing institutions are able or willing to manage AI effectively.

So, how are we to navigate the inevitable rise of a technology that people do not trust, managed by institutions in which they have lost faith?

We face a choice. We can repeat the mistakes of the social media era: political paralysis, rapid deployment, unanticipated consequences, public backlash. This path led to a deeper legitimacy crisis, constituents feeling ignored, the government looking incompetent, and companies operating with reduced social license.

It doesn’t have to be this way. Public concern is not an obstacle; it is the key to building a new form of representative AI governance. Instead of dismissing public anxiety, we can channel it to shape AI development, which will align it with society’s expectations, ease anxiety, and boost adoption. We can shift from a model where technology is imposed upon a skeptical public to one where it is co-created, better representing the people it is intended to serve.

More provocatively, AI itself can be a tool for this governance shift. AI-enabled experiments showcase how the technology can make public engagement more representative, sophisticated, and nuanced, giving way to new consultative methods capable of sensing, aggregating, and acting upon public opinion at a scale previously impossible.

By turning to the public, representative AI governance offers an off-ramp from the increasingly sour public discourse about AI and helps shape a more productive policy path. And, done right, it offers a broader blueprint for how governments and companies can rebuild legitimacy that has been lost.

Public voice guiding the way

Polling is clear that most people want to see policy action on AI, but few agree on what to prioritize first and who should be responsible for what. It’s easy to see how the sheer scale and complexity of issues surrounding AI—from data privacy, cybersecurity, existential risk, social harms, and economic disruption, just to name a few—combined with industry special interests and a lack of AI expertise in government, can quickly result in policy paralysis. Policymakers often seem locked in a game of cat and mouse with technology, as evidenced by Colorado’s repeal and replacement of its landmark AI law.

Governments and companies need not do this alone. They should turn to the public—the very people affected by the technology yet often excluded from shaping it— to help guide the way forward. The public’s concerns about AI can offer a roadmap of sorts: helping policymakers spot, prioritize, and frame issues of societal consequence, as well as help inform solutions.

When ChatGPT burst into the public scene in early 2023, there were a few scattershot efforts to build public input into policy discussions. Yet, many of these pioneering initiatives were seen as “nice to have” and ultimately easily siloed away from core governance. What has changed since these early experiments is that public anxiety has reached a boiling point, amounting to what the Wall Street Journal has termed an AI rebellion. This concern is not just a public relations issue — it is a commercial risk for companies, a political liability for governments, and a threat to adoption.

Now, a growing set of efforts is already underway that offer a path for more systematic representative AI governance. From citizen assemblies like the Snohomish Civic Assembly on AI and the Utah Common Ground Panels on AI to the Collective Intelligence Project’s Alignment Assemblies and Engaged California, different forms of public engagement around AI are emerging, all geared toward constructively channeling public input into policy. When paired with the growing ecosystem of AI tools–like Pol.is, Talk to the City, Remesh, GoVocal, Make.org, and a whole host of other platforms— they can make it easier for people to participate, bridge across even polarized community groups to find areas of consensus, and offer more insights than traditional town halls or focus groups.

This isn’t just about “feel-good” public engagement. Evidence shows that institutionalizing public input yields superior policy outcomes while systematically rebuilding fractured public confidence. For example, Taiwan’s former Digital Minister Audrey Tang found that their governance innovations, which combined the launch of new digital tools with shifts in values and communication, boosted citizens’ trust from a low of 9% in 2014 to 70% in 2020. This builds on decades of research that indicate that deliberative mechanisms like citizens’ assemblies have spillover effects that include improved trust in governing institutions. By turning to the public, policymakers can shape better policy and rebuild trust with their constituents.

This model need not be limited to government. During my time at Meta, we partnered with Stanford’s Deliberative Democracy Lab to build one of the most ambitious public engagement programs by any private company – these deliberative engagements have expanded into a cross-industry consortium effort to ask people around the world how AI companies should design AI technologies. Anthropic has also engaged the public on projects like its Collective Constitutional AI research.

Representative AI Governance is not a replacement for expertise, but instead channels experts to focus on the issues most critical to the public. In this governance arrangement, the public defines the values and the priorities and experts provide the technical implementation. For example, public input can help inform and scope legislation, but experts would then take that signal and implement it in a way that is technically feasible. By empowering the public to shape the policy agenda, we liberate experts to solve technical problems and manage issues beyond public view (e.g., national security) without constantly fighting political fires. In other words, representative governance does not undermine the role of experts but rather complements it by identifying and prioritizing issues of public concern.

Done right, over time, representative AI Governance could help lay the groundwork for a new and more holistic technology governance ecosystem. Over the past 6 years, Meta’s Oversight Board has served as a vital proof of concept in novel technology governance, demonstrating that independent, expert-led bodies can successfully adjudicate complex issues where companies struggle to self-regulate. Now, public consultation could help inform a new set of institutions, bridging the jurisdictional gap that currently separates government action from corporate implementation. One such idea could be the “People’s Constitutional AI Council,” which could move from public input toward public oversight.

Toward representative AI governance

A former colleague once posed a question that has stayed with me: “What if AI happened with people, not to them?” I think this should be our guiding principle for the AI era.

Imagine a world in which our seemingly endless strife about the costs and benefits of technologies wasn’t about people imposing the technology on users. Instead, the public recognizes the very real tradeoffs associated with a technology’s development—and perhaps even embraces them—because they have participated in shaping the design.

Representative AI governance offers governments, companies, and society two benefits. First, it transforms AI governance itself, enabling us to move beyond policy paralysis and technocratic frameworks to input that brings AI alignment closer to the needs and values of real people. Second, it offers a path for broader democratic renewal by confronting what scholar Marjan Ehsassi has termed "voice insecurity" –- the pervasive sense among citizens that their perspectives are no longer heard in policymaking. Institutionalizing deliberative and consultative processes would foster long-term public trust by involving citizens in shaping the technology that impacts their lives.

This vision is within reach and the tools already exist. From deliberative convenings to AI-enabled consensus-building platforms, the infrastructure for representative governance is ready for systematic use. The only missing element is the political will to trust the people. We must act now to ensure that the future of technology is shaped by the society it serves.

By renewing public input across government and industry, we don’t just make AI better – we transform one of the thorniest policy issues into a framework for the future of governance. This is the promise of representative AI governance: an approach that both empowers the public to be an architect in one of the world’s most transformative technologies and guides a broader democratic renewal.