Smart cities bet on AI efficiency, governance still catching up

The study identifies responsible AI as being most visible in conceptual and policy-oriented research, addressing issues such as algorithmic transparency, accountability, bias, and trust. These discussions emphasize the need for explainable decision-making, inclusive design, and ethical oversight. However, they are often disconnected from the technical architectures that drive real-world circular economy applications.

Cities are rapidly integrating artificial intelligence (AI) into sustainability and efficiency strategies, betting on data-driven systems to modernize urban infrastructure. The pressure now is to ensure those systems strengthen trust and equity rather than trade accountability for performance as cities pursue circular economy goals.

A new peer-reviewed systematic review published in Sustainability confronts this tension head-on. Titled “Responsible or Sustainable AI? Circular Economy Models in Smart Cities,” the study sheds light into how responsible artificial intelligence and sustainable artificial intelligence are being integrated into circular economy frameworks within smart-city systems. The research finds rapid technological progress alongside persistent gaps in governance, policy alignment, and ethical integration.

AI powers the circular city but concentrates on performance

The study documents a sharp acceleration in research and deployment of AI-enabled circular economy models across smart cities worldwide. Publication trends show sustained and statistically significant growth, signaling that AI-driven circularity has moved from exploratory research into a mature and influential field. Energy systems, mobility and transport, waste and resource recovery, and smart buildings emerge as the dominant application domains where AI is shaping urban sustainability strategies.

Energy systems sit at the core of this transformation. Predictive analytics and deep learning models are widely used for demand forecasting, smart grid management, renewable energy integration, and optimization of energy flows. These applications play a central role in decarbonization strategies and are strongly linked to overall research influence and citation impact. Mobility systems follow closely, with AI supporting traffic optimization, transport planning, and emissions reduction through data-driven decision-making.

Waste management and resource recovery represent another major pillar. AI-based computer vision, sensor networks, and material flow analytics are increasingly deployed to automate waste sorting, improve recycling efficiency, and support circular manufacturing. These systems are designed to reduce landfill dependence and recover value from waste streams, aligning directly with circular economy principles. Smart buildings also feature prominently, with AI-enabled energy management, lifecycle monitoring, and digital twin integration supporting resource efficiency across construction and operation phases.

Deep learning dominates the technical landscape, accounting for the largest share of AI techniques used in circular smart-city applications. Its strength lies in handling complex, high-dimensional data for image recognition, forecasting, and pattern detection. Reinforcement learning shows the fastest growth rate, particularly in adaptive control applications for energy balancing and traffic systems, although it remains less prevalent than deep learning in deployed systems. Classical machine learning and optimization-based methods continue to play important roles, especially where interpretability and computational efficiency are required.

Despite this technical sophistication, the study finds that AI deployment is overwhelmingly performance-driven. Efficiency gains, optimization outcomes, and system resilience dominate design priorities. While these outcomes are critical for sustainability, they often operate independently of explicit ethical frameworks. As a result, sustainable AI in practice frequently focuses on environmental performance without fully integrating responsibility-oriented safeguards such as transparency, fairness, and accountability.

Responsible AI lags behind operational deployment

The study compares responsible AI and sustainable AI within circular smart-city models. The findings reveal a clear asymmetry. Responsible AI research is concentrated in governance, ethics, and socio-technical discussions, while sustainable AI is embedded primarily in operational systems focused on efficiency and optimization. This separation creates a governance gap in which powerful AI tools shape urban systems without sufficient integration of responsibility principles.

The study identifies responsible AI as being most visible in conceptual and policy-oriented research, addressing issues such as algorithmic transparency, accountability, bias, and trust. These discussions emphasize the need for explainable decision-making, inclusive design, and ethical oversight. However, they are often disconnected from the technical architectures that drive real-world circular economy applications.

On the other hand, sustainable AI is operationalized through energy optimization, waste automation, mobility analytics, and digital infrastructure. These systems deliver measurable environmental benefits but may lack explicit governance mechanisms to address data privacy, institutional accountability, or social equity. The result is an unaligned zone where high-performance AI operates without clear responsibility framing.

This gap is not merely theoretical. The study’s knowledge-translation analysis shows that broad alignment with sustainability goals does not reliably translate into policy adoption or governance integration. While many AI-driven circular economy initiatives reference Sustainable Development Goals, the depth of policy implementation remains uneven. Research intensity alone does not ensure regulatory uptake or institutional embedding.

The analysis highlights data interoperability and institutional fragmentation as major barriers to translating AI innovation into policy outcomes. Fragmented governance structures, inconsistent data standards, and weak coordination between public and private actors limit the ability of cities to institutionalize responsible AI practices. Even where technical capacity is high, governance mechanisms often lag behind.

City readiness shows governance makes the difference

To illustrate how these dynamics play out in practice, the study examines city-level models of AI readiness, circular economy integration, and policy alignment. The findings show significant variation across cities, underscoring that governance coherence rather than technological intensity determines successful integration of responsible and sustainable AI.

Cities such as Singapore, Amsterdam, and Helsinki emerge as balanced leaders. They combine high AI readiness with strong circular economy integration and coherent policy frameworks. In these contexts, AI deployment is supported by institutional mechanisms that align technical innovation with ethical oversight and sustainability objectives. Public-private collaboration, coordinated data governance, and proactive policy design enable these cities to translate research and experimentation into systemic transformation.

Other cities demonstrate high technical capacity but weaker governance alignment. In these cases, AI systems are deployed extensively in energy, mobility, or waste management, but policy integration remains partial. The study shows that without strong institutional embedding, innovation risks remaining fragmented, limiting long-term impact and public trust.

The analysis reinforces the role of collaboration as a critical enabler. Public-private partnerships, knowledge co-creation, and cross-sector coordination significantly improve translational efficiency. Conversely, lack of interoperability and institutional silos undermine the ability to scale responsible AI practices. Governance asymmetry persists even as technical maturity increases, highlighting the need for deliberate policy intervention.

Overall, the study asserts that responsible and sustainable AI cannot be treated as interchangeable concepts. Sustainable AI focused on efficiency and environmental performance is necessary but insufficient. Without responsible AI governance, circular smart cities risk prioritizing optimization over inclusivity, accountability, and legitimacy.

Bridging innovation and governance in circular smart cities

AI-driven circular economy transitions are at a crossroads. Technological capability has advanced rapidly, and AI systems now play a central role in managing urban energy, mobility, waste, and infrastructure. However, governance frameworks have not evolved at the same pace.

The research argues for policy-responsive AI governance that embeds responsibility principles directly into system design and deployment. This includes integrating ethical oversight into urban planning, establishing clear accountability mechanisms, harmonizing data governance standards, and ensuring citizen engagement in decision-making processes. Responsible AI is framed not as a constraint on innovation but as a condition for sustainable and inclusive transformation.

For policymakers, the findings highlight the need to move beyond compliance-based regulation toward anticipatory governance models that align innovation with public values. For technologists and industry actors, the study underscores the importance of designing AI systems that support transparency, interoperability, and lifecycle accountability. For cities, it reinforces that smartness is not defined by technological intensity alone, but by the capacity to govern complexity responsibly.