SafeCity: Smart Surveillance and Crime Prevention

Authors | Anbisha Humagai Krishna Budhathoki Tisa Kasaju Amol Dangol Prashanna Raj Singh

The adoption of modern technologies and its usage in the city has increased the rate of implementation of the concept of smart city surveillance systems intended to enhance the crime prevention citizens safety. SafeCity is an example of a platform that integrates artificial intelligence, computer vision, IoT devices and big data analytics to track and analyse activities in cities. These technologies allow law enforcement to see the possible risks and distribute policing resources effectively by connecting CCTV cameras, facial recognition, gunshot detection, automatic number plate recognition system, and predictive policing algorithms. Using these systems raises worries about whether AI is making fair and responsible decisions, especially since algorithms can sometimes be biased. Moreover, the mass acquisition of personal data raises the issue of cybersecurity and privacy, which need powerful protection and control through regulatory measures. In addition to technical challenges, there exist other implications in society, such as the effects on civil liberties, economic investment, sustainability because of high energy usage, and the necessity of proper governance and research to allow responsible technological development.

Responsible AI and Ethics

Artificial intelligence-based surveillance systems like SafeCity raise important questions regarding responsible AI and ethical responsibility majorly in terms of algorithmic discrimination and its consequences. The principles of responsible AI state that automated systems should work fairly without causing any discrimination to any particular groups. Research into biased facial-recognition systems indicates that misidentification of women and people with darker skin tone is higher due to prevalent biased datasets (Tavares, 2023). In the context of law-enforcement, the deployment of such systems may result in misidentification, unjustified police investigation, and oversight of minority groups. Ethical issues are also reinforced by the existence of predictive policing technologies that are heavily dependent on past crime statistics. Machine-learning systems replicate patterns found in their training data, which caused neighborhoods that were disproportionately focused in the past to be still classified as high-risk reinforcing the same pattern (Zhou et al.,2024). Due to this, the existing social inequalities are further exacerbated rather than providing fair evaluations of crime. To mitigate bias, responsible research and development practices must be integrated into AI system design including fair algorithm design, diverse training datasets and systematic bias testing before deployment. It is essential to integrate such precautions into the development lifecycle to ensure that AI surveillance technologies are in line with the professional ethics practice and are not unintentionally used to propagate discrimination.

In addition to algorithmic bias, transparency, governance structures and technical reliability are other major factors to be considered in case of surveillance technology. The adoption of such facial recognition-based systems heavily relies on the trust of people in institutions and the views of fairness in the application of the technology. Empirical evidence regarding the perceptions of people towards these systems in policing indicate that the level of trust in governmental bodies is the major factor influencing the decision to approve or reject AI-driven surveillance systems (Li, 2024). In order to build trust, authorities and developers should be transparent about how decisions are made and must require proper accountability if any wrong decisions are made. Facial-recognition technologies also pose serious ethical issues in their operation under technical performance constraints in the uncontrolled environment like the urban surveillance networks. Studies on the effectiveness of facial recognition in law enforcement reveal that reduced image brightness, including motion blur and low-resolution images may significantly increase incorrect identifications and an excessively large share of particular demographic group (Cuellar et al.,2025).To address these issues, recent guidelines of the IEEE state that structure of evaluation frameworks of biometric technologies is important , with performance testing methods, which are aimed to assess accuracy, reliability and bias of facial-recognition systems, being developed as a response(IEEE, 2023). These principles highlight the need of strict technical assessment and policy controls in accountable use of AI based surveillance technologies. The interdisciplinary research, promoting transparency in evaluation and system design that prioritizes fairness can help developers and policymakers reduce the level of algorithmic bias and build more trust among people in the smart-city surveillance systems.

Cybersecurity and Privacy

Although SafeCity was marketed as a tool to help reduce crime, its vulnerabilities in cybersecurity posed risks that were damaging at the system level. The system was successfully hacked and the real-time location data of thousands of citizens and undercover police officers was leaked due to the implementation of  interconnected surveillance infrastructure without adequate security measures. Ali et al.(2025) highlight that the use of AI and IOT in smart city networks leads to an increase in the potential attack surface so much that the traditional security methods revolving around the perimeter become completely insufficient. Since thousands of edge devices continuously contribute data into centralised databases, SafeCity created exactly the sort of high value targets that are currently the focus of modern adversaries to inflict harm. The sensitivity of data involved compounds this risk considerably; real time location feeds for undercover officers represent a direct threat to operational safety, not merely a digital inconvenience. Kwon, Salim and Park (2023) observe that smart city security failures frequently arise not from weaknesses in individual components, but from poor integration between subsystems, across its camera networks, ANPR systems, and gunshot detection sensors. Zeng et al. (2024) argue that AI-based anomaly detection frameworks should be implemented at the network level to identify intrusions before data exfiltration occurs, yet SafeCity evidently lacked this proactive monitoring capability entirely, leaving the system open to exploitation long before the breach was discovered.

SafeCity has a major violation of its basic privacy principles. The system continuously collects biometric data at scale, facial geometries, vehicle movements, and precise citizen location traces yet it has no legal framework that specifies retention periods or access controls. Zhao (2023) claims that a well functioning governance system is a prerequisite for management of data resource AI-powered smart cities. However, SafeCity does not comply with any of these criteria. Izonin, Chretien and Ebrahimnejad (2022) argue that trustworthy AI systems should be programmed with privacy-by-design rather than the data protection regarded as something to do after deployment, which is exactly what SafeCity did, focusing on operational capabilities at the expense of the rights of the citizens. Addressing these failures requires solid measures on three separate levels. SafeCity should be moved to a federated architecture where surveillance data stays on the edge devices and is only processed locally, rather than being collected on centralized servers. An independent cybersecurity audit carried out by persons completely unassociated with the Metropolitan Police or SafeCity has to cover all network vulnerabilities, including the used encryption standards, and access control policies. And, Parliament should pass legislation that would provide clear statutory guidance which stipulates binding retention time periods and hierarchical access control systems structure for the biometric surveillance data. Lack of these measures means that the crime-reduction data reported by SafeCity could not be used as moral grounds to justify the extensive damage caused through its security lapses not only to the citizens but also to the police officers.

Sustainability and Energy Usage of Smart Surveillance Systems

SafeCity systems combine artificial intelligence, computer vision, and connected sensors to monitor urban environments and improve public safety. While they can help law enforcement detect crime faster and respond more efficiently, they also raise questions regarding 9 environmental sustainability and social impact. Understanding these effects is important to ensure that technological progress does not create long-term harm for society or the environment. One of the main sustainability concerns of AI-powered surveillance systems is their high energy consumption. Systems like SafeCity operate continuously, analyzing thousands of video feeds and processing large datasets in real time. This requires powerful computing infrastructure, including edge devices and large data centers that run complex machine-learning models. According to the International Energy Agency, data centers used for digital technologies and AI consumed between 240 and 340 terawatt-hours of electricity globally in 2022, showing how energy-intensive digital infrastructure has become (International Energy Agency, 2023).When smart surveillance networks expand across multiple cities, the energy demand can increase significantly, disrupting long-term environmental sustainability.

The environmental impact of AI systems also extends beyond electricity use. Large data centers require cooling systems and physical infrastructure that contribute to carbon emissions and water consumption. Research on sustainable artificial intelligence shows that the rapid growth of AI technologies could significantly increase environmental pressure if energy-efficient computing systems are not implemented (Vinuesa et al., 2021). This means that governments and technology companies must consider greener alternatives such as renewable energy, efficient data processing methods, and optimized AI models when deploying large-scale surveillance networks. In addition to environmental concerns, smart surveillance systems can have important social consequences. Ramli, Azizi and Thurairajah (2024) state that smart technologies in cities can be classified into human–technology interaction, technology interaction, and energy system management layers. The study reviewed 60 articles and identified 92 smart technologies related to energy consumption. These were classified into three layers: human–technology interaction (HTI), technology interaction (TI), and energy system management (ESM). Smart grids were the most discussed in ESM, while HTI included household technologies, and TI involved IoT, sensors, cloud systems, and big data. Overall, AI-powered surveillance systems like SafeCity present a complex balance between safety benefits and sustainability challenges. Their high energy requirements contribute to environmental concerns, while widespread monitoring can influence public behavior and social trust. For these technologies to be sustainable in the long term, governments and technology developers must prioritize energy-efficient design, transparent governance, and respect for citizens’ rights.

Social impact and economic factors

The increasing adoption of smart city technologies has transformed the approach that modern urban environments use to address public safety and crime prevention. On the surface, developments in policing that utilize more accurate crime predictions appear promising for enhancing police response effectiveness. However, challenges to their legitimacy arise due to the risks associated with Big Data Predictive Policing (BDPP) (Lee, Y., Bradford, B. and Posch, K. 2024). From a social perspective, the expansion of large-scale surveillance infrastructures can reshape the relationship between citizens and the state. Continuous monitoring of public spaces may create feelings of reduced privacy and increased scrutiny, potentially altering how individuals behave in public environments. Community groups have expressed concerns that constant observation can discourage participation in protests, social gatherings, or political demonstrations due to fear of monitoring. Additionally, algorithmic policing systems may reinforce existing social inequalities if predictive models rely on historically biased policing data. This can lead to disproportionate targeting of certain neighbourhoods or demographic groups, increasing social tensions and reducing public trust in law enforcement institutions (Khan, F. 2025). While governments often justify these costs through potential reductions in crime and improvements in urban safety, the financial burden of implementing and maintaining such systems raises questions about long-term sustainability and the allocation of public resources within smart city development strategies.

Adding on, safer urban environments may also encourage economic growth by attracting businesses, tourism, and investment into areas perceived as secure and well-managed. Also, the development and deployment of smart surveillance technologies can stimulate economic activity within the technology sector, creating demand for data analysts, cybersecurity specialists, and infrastructure engineers who support these systems. However, the broader social consequences of large-scale surveillance infrastructures present significant challenges that must be carefully addressed. Continuous monitoring of public spaces may create a “surveillance culture” in which citizens feel constantly observed, potentially limiting freedom of expression and discouraging civic engagement. Concerns regarding algorithmic bias and unequal facial recognition accuracy may lead to unfair treatment of certain demographic groups, contributing to social inequality and distrust toward law enforcement agencies (Arooj Sikandar, Bibi, A. and Associate, S. 2024). The economic implications of maintaining sophisticated surveillance systems, including energy consumption from edge computing technologies and the requirement for effective cybersecurity solutions, raise concerns with regard to long-term economic viability as well. Thus, while smart surveillance technologies can play an important role in improving public safety, their implementation must be balanced with technological progress, social responsibility, economic viability, and the preservation of basic democratic values.

Code of Conduct

The BCS code of conduct is a fundamental normative text that presents a set of professional responsibilities of computing practitioners all over the world. There are four major duties on which it is constructed. First, a professional must put the interest of people at centre and have the work of practitioners ensure that the work adheres to the overall wellbeing of the society. Second, competence requires practitioners to take up only work that they are competent to do, and expand their knowledge by developing as professionals. Third, integrity involves complete honesty in any professional interaction such as full disclosure of conflicts of interest and avoidance of deceiving clients, employers, and general population. Fourth, it requires the practitioners to safeguard the reputation of the computing profession and always behave within the law. All these responsibilities are enforceable duties which have real disciplinary consequences in case of nonadherence. There are indications that codes of ethics do not yield any serious results when organisations passively assume that they are aware of them by being passive, but by training people (Kluge Correa et al., 2025). It also stipulates that professionals must critically analyse social implications of their practice more broadly, by rejecting systems of discrimination or undermining democratic accountability (Laux, Wachter and Mittelstadt, 2024). The cross-jurisdictional analysis of professional codes in information systems additionally attests to the fact that the primary responsibilities of public interest, competence, and integrity are evident in all institutional and legal settings (Ribeiro and Varajao, 2025).

Research & Development

The application of SafeCity research and development is a growing field of technologies, incorporating various AI based systems. The advantages of these technologies are significant, and efficiency in responding to an emergency, distributing policing resources, and situational awareness of city management authorities have been improved. Regardless of these benefits, computing professionals have some severe ethical obligations that they cannot escape. SafeCity algorithms are observed to reproduce racial and socioeconomic disparities, and predictive policing algorithms increase the over-policing of marginalized communities because they base their training data on historically biased data (Hung and Yen, 2023). The scale of biometric and location data collection is also leading to additional issues of function creep when the data collected in a given application purpose has subsequently been used in other areas outside its original consent purpose, breaching the duty of data protection and the BCS duty of integrity. As a result, the Code entails that professionals engaged in SafeCity R&D must champion privacy-by-design strategies that include data minimization and purpose limitation at the very beginning of development. By rendering many AI systems less transparent, democratic accountability also becomes a problem because the principle of transparency by the BCS system makes practitioners encourage explainability and make automated decisions made on behalf of the masses challenging to scrutinize (Mittelstadt, Wachter and Russell, 2024). Ethical evaluations have to be instilled in the R&D process with the involvement of disciplinary cooperation such as law, sociology, and community engagement, to make sure that the SafeCity technologies are in the interest of the people.

The technologies of smart cities surveillance prove the opportunities and challenges which are connected with the implementation of highly developed digital systems in the sphere of crime prevention. Although such systems have the potential to improve the efficiency of the law enforcement process and make cities safer, the introduction of such systems should take into account more universal ethical, social, and economic factors. Meanwhile, strong cybersecurity measures and privacy rights are needed to make sure that sensitive biometric and location data is not abused or targeted by cyber attacks. The issue of diminished civic freedom and worries about mass surveillance are some of the social effects of this, and hence the need to uphold a sense of trust among the people. Economically, the governments should consider expenses of infrastructures, maintenance and energy usage whilst weighing the possible good results like increased security and economic confidence. Strong governance, regulation policies and future research and development are thus required to see that smart surveillance technologies are applied in a responsible and sustainable way.

This research report has been prepared by students of The Westminster College, Kupandole, as part of their module requirements. Students are required to write a research paper and publish it on their LinkedIn profiles.