AI will Transform Cities. Can Cities Transform AI?
Jinhua Zhao, J. Phillip Thompson, Ross Gittell, Michael Leong, and Kevin F. Hsu
3. How do cities capture and share the value that AI promises to create?
In this section, we outline the key actions cities must take at the interface of the public and private sectors to ensure the broader societal benefits. This is in contrast to the previous section, which focused primarily on internal public sector applications of AI. Here, the emphasis shifts to the broader ecosystem in which cities operate, while still squarely within the city’s key levers and sphere of influence.
We believe that municipal governments must pursue three key areas at this interface: increasing human capital, improving physical infrastructure, and developing public trust. These key actions address a larger question: will the benefits that we see from AI, either in the public or private sector, translate into shared prosperity, expanded human capabilities, and strengthened community well-being? Or will they instead accelerate inequality through workforce displacement, uneven access to technological benefits, and insufficient public-sector capacity? These specific priorities have policy implications in both an internal municipal capacity and a broader societal capacity. Table 4 below distinguishes how each area operates across different levers of city influence:
Area 1: Knowledge of AI
Strengthening a city’s capabilities in human capital includes building multiple levels of AI competence across the population. At a baseline level, residents should possess enough familiarity to interpret AI outputs critically, recognize limitations, and guard against misuse such as scams or misinformation. At a professional level, workers should be able to use AI tools effectively in their daily tasks to enhance productivity and job quality. At an executive level, leaders must be able to make informed decisions about how to integrate AI alongside human labor, balancing efficiency gains with workforce impacts. Finally, at a regulatory level within government, officials must understand how AI systems function, their risks and safeguards, and how to design policies that protect the public interest. Figure 6 illustrates these four distinct levels of competence.
At the Baseline level, governments should aim to equip everyday constituents with basic AI knowledge through public education and public engagement campaigns. Informed citizens will be empowered to use AI-enabled public apps, recognize the power and limitations of AI, and better distinguish AI-generated scams or misinformation. These skills are particularly salient for vulnerable populations such as elderly, low-income, or non-English speaking residents, featuring some overlap with baseline digital literacy.
At the Professional level, governments can muster the capacity to retrain and upskill large numbers of workers, where the nature of their jobs is likely to be changed by AI, so the entire city’s workforce remains competitive and adaptable. Cities can explore partnerships with local educational institutions or private companies to scale continuing education, or offer incentives to local employers to retrain workers. Within the municipal government’s own workforce, many public employees must also be trained to this level.
At the Executive level, leaders in both the public and private sectors must be prepared to guide organization-wide transitions that integrate AI with human labor. Governments can develop this capacity by investing in specialized training for senior administrators and establishing clear public-service principles that may differ from private-sector practices, while the private sector is likely to develop this capability on their own.
At the AI Governance level, governments should cultivate a smaller group of specialists with deep technical and policy expertise to understand AI systems and their potential social impacts. These individuals would be responsible for setting procurement standards, conducting audits and evaluations, and ensuring that public-sector AI deployments remain fair, accountable, and aligned with the public interest. While this capacity would primarily focus on government uses of AI, it may also extend to a watchdog role over private-sector applications in critical domains such as housing, transportation, and employment, with the authority to investigate harms and pursue legal action in cases of malpractice.
Policy Examples: Building Human Capital and Workforce Training
Finland: Digital Compass Strategy
Since 2022, Finland’s digital compass strategy has cultivated a digital “building” culture, where digital literacy is defined as a core life skill and part of one’s ethical and civic responsibility. Through targeted programs such as municipal education centers, tripartite cooperation, using libraries as digital access points, and targeted outreach to at-risk groups such as rural or low-income households, basic digital literacy in Finland has reached over 80%. While digital training does not address AI specifically, this positions the country and its workforce well for increased AI adoption at broad-based levels.
Singapore: SkillsFuture Career Transition Program
The SCTP provides training and placement opportunities for mid-career professionals in Singapore, with a focus on 3-12 month courses at accredited institutions (such as local universities), placements, and career advisory. All Singaporeans are given individual SkillsFuture credits to use towards these courses, with additional subsidies available based on need. As of 2024, over 100,000 workers use SkillsFuture credits for mid-career courses annually, with these courses being a combination of AI and non-AI related courses. Several courses also come with a 6-month free subscription to premium AI services.
UK: Government - Private Industry AI Skills Partnership
The UK has partnered with top technology companies such as Amazon, Google, and Microsoft to provide AI training with the ambition of equipping 7.5 million workers, or about 20% of the UK’s workforce, with AI skills by 2030. These trainings will focus on understanding how AI systems such as chatbots and large language models work, and how they can be used for sector specific practical applications. Separately, all civil servants from England and Wales will undergo similar AI training from the Fall of 2025.
UAE: Government AI Training
Since 2018, the UAE AI Office has offered advanced AI courses for government employees, focusing on skills needed to become AI ambassadors in their fields, typically in year-long courses with capstone projects. The plan aims to ensure that the entire senior government leadership, including ministers, are well versed in AI, with lower level government employees receiving more ad-hoc training. In addition, the UAE Office of AI conducts an annual Summer AI camp where government agencies and tech companies learn about AI tools, ethics, and applications are conducted together with students and the general public.
MIT: Day of AI
Open curricula such as Day of AI, an initiative from the MIT Responsible AI for Social Empowerment and Education initiative, provides teaching materials that train K-12 students on basic AI fundamentals such as how computers represent information, how predictive and generative AI works, and practical use cases such as social media recommendation algorithms. (dayofai.org/curriculum)
Area 2: Physical & Digital Infrastructure
It is also imperative that a city strengthen its infrastructure (physical capital) in anticipation of AI adoption. While infrastructure gaps have long underpinned the digital divide, expanded AI use risks widening these disparities if access to connectivity, computing resources, and modern public facilities remains uneven. Cities must therefore build capacity in core areas, including municipal data systems, secure application and cloud infrastructure, data transmission networks, equitable public access to digital services, and incubation interfaces that allow AI solution providers to engage responsibly with city agencies.
At the same time, cities do not need to develop fully vertically integrated AI infrastructure to benefit from these technologies. Many urban AI solutions will be delivered by non-government actors, including established technology firms, local startups, non-profits, university initiatives, and public–private partnerships. These actors require a capable municipal partner that is ready to engage, share data responsibly, and align solutions with public objectives. Accordingly, rather than cities overemphasizing hardware ownership, they are better served by focusing on data readiness, process readiness, broad accessibility, and fostering a civic AI ecosystem that can develop, scale, and deploy solutions in the public interest. Figure 7 highlights four arenas where governments can upgrade infrastructure and public spaces to support responsible public-sector AI deployment. Figure 7 highlights four areas where governments can upgrade infrastructure and public spaces to support responsible AI deployment, with a focus on public sector applications.
Firstly, cities should ensure data availability by building infrastructure to assemble, store, and maintain high-quality municipal data that can support AI applications across both public and private use cases. This includes establishing clear protocols for data collection, updating, security, and anonymization of personally identifiable information, as well as distributing datasets through controlled interfaces such as APIs. While similar in intent to existing open data initiatives, AI-enabled applications require more advanced data governance, quality assurance, and technical operations to ensure reliability and responsible use.
Secondly, cities should develop secure deployment capacity by establishing infrastructure to host, store, and transmit outputs generated by AI systems in a safe and consistent manner, whether through on-premise or cloud-based environments. A unified deployment environment with strong security standards and clear operational protocols can better protect sensitive government data, improve efficiency, and enable coordinated incident response across agencies.
Thirdly, cities should invest in infrastructure for universal public access to ensure that residents can benefit from AI-enabled public services and economic opportunities. This includes regulating existing broadband markets to prevent price gouging and improve affordability, exploring municipally run broadband as either a competitive alternative or a public utility, and expanding public internet access points in libraries, schools, and other civic facilities. Cities may also consider providing AI subscriptions or compute credits to support local businesses with less resources. Together, these measures help close persistent connectivity gaps, prevent AI adoption from deepening existing digital divides, and ensure that vulnerable populations can meaningfully share in the gains from AI development.
Finally, cities should create incubation spaces for AI solutions that foster a civic-oriented ecosystem of developers, startups, researchers, and community partners. These spaces—whether physical or virtual—serve as controlled testing environments where new tools can be piloted, evaluated, and refined before deployment in live public settings. By providing a safe arena for experimentation, engagement sessions, and collaborative workshops focused on priority municipal use cases, cities can reduce the risk of premature public rollouts and ensure that AI systems are validated and aligned with local needs and public objectives before full-scale implementation. Physical spaces could be created by repurposing government offices or co-locating programs within museums, libraries, or other civic buildings, while virtual spaces could take the form of secure sandboxes, shared data environments, or online collaboration platforms that allow partners to test and iterate on applications using municipal datasets.
Policy Examples: Building Physical and Digital Infrastructure for AI Deployment and Access
UK: Creating a National Data Library
The UK’s AI Opportunities Action Plan mandates the creation of a National Data Library tasked with identifying high-impact public datasets and strategically shaping what data can be collected to train AI models. Additionally, they are tasked to finance the creation of new datasets that can meet the needs of AI researchers and companies, as well as establish copyright-cleared British media training datasets.
Japan: Deploying Data Infrastructure for AI
Japan’s National AI Strategy calls for the establishment of data collaboration infrastructure for critical industries such as healthcare, transportation, disaster resilience, and regional revitalization, as well as processes to ensure the quality of big data. In addition, the plan calls for the nationwide installation of 5G mobile communications and ensuring safety and reliability so that AI can be used throughout Japan.
Singapore: AI Placemaking and National Supercomputing Center
As part of Singapore’s National AI Strategy, an iconic AI site was developed to “co-locate creators, practitioners, and nurture the AI community in Singapore”. The site hosts regular events and shared activities, including networking events, support hours for usage of government AI tools and datasets, workshops, and other community driven activities.
Singapore’s National Supercomputing Center also provides free (limited) GPU access to every citizen and discounted rates for approved projects submitted from local universities and research institutes.
US: Municipal Broadband
Over 400 US cities have installed municipal or community owned broadband networks which provide high-speed internet access for affordable rates as low as $10 a month, compared to the national average of over $70 a month. These networks are often operated directly by city utilities or contracted to non-profits, with the goal of expanding digital access and promoting competition with private operators.
Area 3: Public Trust & Civic Consciousness
Beyond strengthening human capital and physical infrastructure, cities must also cultivate public trust and civic consciousness around the use of AI. Even the most technically sound systems will face resistance or fail to achieve their intended benefits if residents do not understand how they work, why they are being used, or how their rights and interests are protected. Building trust therefore becomes a core component of AI readiness, requiring transparent communication, meaningful public engagement, and clear accountability mechanisms. This section turns to this third area, examining how cities can foster informed civic awareness and sustained public confidence in AI deployment.
Table 5 classifies the potential impacts and harms of AI into three categories: primary effects on individual users, secondary effects on social and economic structures, and tertiary environmental effects.
Governments are generally more experienced in managing tertiary impacts, as these resemble externalities commonly addressed through planning, permitting, and environmental regulation. However, they may have less institutional capacity or expertise to anticipate and govern the primary and secondary impacts of AI, which involve individual rights, labor markets, and broader social equity concerns. For this reason, we recommend that governments engage public stakeholders in three complementary capacities.
-
Establish a Civic AI Board, which would act as a non-partisan advisory body to the government and a tool to mediate societal concerns. This board could comprise representatives from academia, civic institutions, trade unions, the legal profession, non-profit groups, the private sector and the general public. They would jointly advise government agencies by drawing from diverse views about standards of care expected of the municipality, companies, and citizens.
-
Establish partnerships with local stakeholders including civic organizations, private companies, academic institutions, trade unions, and peer governments to develop specific programs for policy and training needs that are identified by the Civic AI Board. In particular, local academic institutions could function both as a qualified neutral third party to advise on acceptable standards for AI application deployment, and also be a provider of workforce training programs.
-
Establish inter-governmental consortiums with peer municipalities to share best practices, pool expertise, and coordinate standards. Neutral third parties, such as academic institutions or nonprofit organizations, could provide a convening platform to facilitate joint research, comparative evaluations of AI systems, shared training programs, and the development of common procurement guidelines that benefit all participating jurisdictions.
Policy Examples: Addressing Risks and Potential Harms of AI
Los Angeles: AI Safety Checklist
The City of Los Angeles' AI safety checklist declares that “AI cannot be relied upon for decision making if it perpetuates bias and harms L.A. communities,” suggesting the need to mitigate these potential direct impacts. The LA AI Safety Checklist includes “safeguards to detect and reduce biases in data, algorithms, and outcomes” including the check for demographic (race, gender, socioeconomic) biases prior to the launch of an AI tool, along with periodic audits of models to “identify, report, and mitigate any biases that emerge after implementation.”
San Jose: Facilitating Model Transparency
San Jose’s AI Policy, outlined in City Policy Manual 1.7.12, emphasizes proactive transparency, requiring that the purpose and use of AI systems be clearly communicated and documented for the public. The policy states plainly that residents will be informed whenever AI is used, and that system data sources, operational models, and governing policies should be understandable and accessible. The city’s AI handbook also allows third-party auditors to access system data for independent evaluation. For generative AI, the city similarly foregrounds transparency to build public trust, requiring disclosure when AI is used and mandating that all outputs be vetted by staff to ensure communications remain factual, credible, and verified.
San Jose: Supporting Equitable Outcomes
San Jose’s AI policy states its intent for AI systems to “deliberately support equitable outcomes for everyone,” with the aim of managing bias “to reduc[e] harm for anyone impacted by the system’s use.” Its Algorithmic Impact Assessment and its AI Handbook pose several critical questions that must be answered, about target audience (general public vs. specific group), impacts on children under 18, and whether an individual’s rights or freedoms, economic status, or health are implicated.
New York City: Partnering Agencies with Allied Specialities
As part of its AI Action Plan, New York City will tap the Department of Consumer and Worker Protection to enforce a local law requiring employers and employment agencies that use “automated employment decision tools” to provide notification, and to conduct and publish “bias audits” of the tools. NYC will also call on the City Commission on Human Rights—a body that enforces the local Human Rights Law, intended to protect residents from bias and discrimination in employment, housing, and public accommodations—to consider how problems could arise with the use of AI tools.
Virginia: Physical Footprint Impacts
Virginia legislators have proposed enhanced environmental and community impact assessments to precede new data center approval. These include assessing local air pollution, sound, and land use impacts, as well as measures that promote fuel switching from backup diesel generators to less-polluting options.
Policy Examples: Community Engagement for Building Trust in AI
Washington DC: Establishing an Advisory Group
Through a mayoral order, Washington, D.C. has created a multi-stakeholder AI Advisory Group responsible for holding public listening sessions to gather community input on government use of AI tools and review specific AI deployments. This advisory group includes government officials and members of the public with expertise in racial equity, disability rights, labor relations, criminal justice, education, healthcare, privacy, cybersecurity, and transparency/open government, reflecting a range of different experiences and perspectives.
San Francisco: The Civil Grand Jury
The San Francisco Civil Grand Jury comprises 19 randomly selected jurors who serve approximately 500 hours over 1 year periods to investigate the operations of various city departments, personnel, or issues of interest to the public. One such issue they were tasked to investigate and make recommendations for was the city’s original policy towards AI use and procurement in the government.
San Jose: Online and In-Person Engagement
San Jose plans to create an interactive online portal for residents to provide feedback on AI initiatives, participate in surveys, and join discussions, complemented by live social media town halls for real-time Q&A. The city will also host public workshops and town halls, partner with local institutions such as libraries to support citizen advisory boards, and conduct focus groups with key stakeholder groups. In addition, interactive kiosks in public spaces like libraries, parks, and government buildings will allow residents to learn about AI and share input, while designated “living labs” in selected neighborhoods will pilot AI technologies and gather real-world feedback.
Lastly, given the significant investment required to build human capabilities, physical infrastructure, and public trust, governments should consider establishing sustainable funding mechanisms that capture some of the economic value generated by widespread AI activity. Potential approaches could include targeted taxes on AI-enabled services, levies on related digital activities, or revenue derived from the responsible use of government data assets. Equalizing the effective tax burden between human labor and algorithmic services may also help mitigate distortions in labor markets and ensure that productivity gains from automation contribute to shared public resources.
While dedicated taxes for technology governance are uncommon in the United States, analogous models already exist, such as tolls used to finance transportation infrastructure, value-added taxes on tourism in hotels and resorts, and local fees imposed on transportation network companies like Uber and Lyft. These precedents suggest that carefully designed revenue instruments could provide stable funding for municipal programs that manage AI’s impacts and support the infrastructure and oversight needed for responsible deployment. Table 6 compares several potential revenue sources that governments may explore to generate long-term, sustainable support for AI governance and infrastructure.
As AI continues to reshape the technological, social, economic, and political environment in which cities operate, these three coordinated investments in human capital, physical infrastructure, and public trust have the potential to significantly amplify municipal capacity. These investments ensure that cities are not merely passive adopters of externally developed technologies, but active participants in shaping how AI is deployed and experienced in urban life. By investing in people, platforms, and public engagement simultaneously, cities can move beyond reacting to technological change and instead play a proactive role in shaping the direction of AI development, aligning its use with democratic accountability, equitable growth, and long-term community well-being.
In conclusion, AI will transform cities; but we firmly believe cities can also transform AI.
This is Working Version 1.2 of the white paper.
We welcome your feedback as this publication continues to evolve.
AI will Transform Cities. Can Cities Transform AI?
Jinhua Zhao, J. Phillip Thompson, Ross Gittell, Michael Leong, and Kevin F. Hsu