When environmental harm becomes a public safety issue: our new AI MAPS use case

Public safety is often imagined through visible events such as a violent incident or a disturbance in public space. Something that happens to someone, somewhere, at a particular moment.

But many of the risks shaping how safe public spaces truly are do not appear in that way. They unfold quietly, cumulatively, and often without a clearly identifiable victim. Pollution entering waterways. Toxic substances dispersing through air. Environmental degradation that does not immediately trigger alarm, yet gradually undermines the conditions that make life possible.

If public safety is understood only through direct harm to individuals, these risks can remain strangely peripheral. If instead public safety is approached as a quality of the environments in which living beings coexist, environmental harm begins to look less like a secondary concern and more like a central one.

This shift in perspective is at the heart of our third AI MAPS use case. After studying crowd control at demonstrations and feelings of safety in residential neighbourhoods, we're now asking: how can AI help prevent and detect high-impact environmental harm?

Beyond human-centred safety

Discussions about safety frequently revolve around perception. Whether people feel safe walking through a neighbourhood. Whether public spaces appear orderly and predictable. These are important dimensions of public safety, and research has shown how strongly perceptions shape behaviour and wellbeing.

Yet perceptions do not always align with underlying risks. A space may feel safe while environmental conditions within it are deteriorating. Conversely, improvements in perception do not necessarily correspond to improvements in ecological health.

This raises a deeper question. If public space is shared not only by humans but by a wide range of living species, can safety be defined solely through human experience?

An ecosystemic perspective suggests otherwise. It directs attention to the biophysical conditions that sustain life across species, regardless of whether those conditions are immediately visible or perceptible. Among these, access to clean air and water stands out as fundamental. When these conditions are compromised, the safety of public space is affected in ways that transcend individual perception.

From this viewpoint, environmental harm can be understood as a form of high-impact harm. Not because it always produces dramatic events, but because it alters the basic conditions upon which multiple forms of life depend. 

Why this matters for public safety and for AI 

In traditional crime policy discourse, "high impact crime" refers to crimes with severe effects on victims: robberies, homicides, explosions. The human victim sits at the center. There's nothing wrong with that. But our previous research taught us that a "good life in public space" involves all species that share that space with humans. Illicit actions that damage the ecosystem affect everyone. 

The examples are everywhere:

• Releasing untreated wastewater into rivers 
• Using pesticides against regulations 
• Poisoning the air that every living thing breathes

When we study AI practices that prevent or detect environmental harm in public space, every species becomes a beneficiary.

Why AI enters the picture 

Addressing environmental harm is not a new societal challenge. Regulatory frameworks, monitoring infrastructures, and inspection practices have long been in place to protect air and water quality. At the same time, contemporary environmental risks are characterised by scale, complexity, and the speed with which conditions can change.

The potential of AI in this context lies less in autonomous decision-making and more in its capacity to assist with pattern recognition, early detection, and the integration of diverse data sources. Environmental sensors, satellite data, inspection reports, and logistical information can collectively form a rich but fragmented picture of environmental conditions. AI techniques can help identify anomalies, detect emerging risks, and support more timely interventions.

Equally important is the possibility of shifting from reactive enforcement towards proactive safeguarding. Instead of responding only once harm is visible, AI-supported systems may contribute to signalling risks earlier, enabling authorities and other stakeholders to act before degradation escalates. 

Working with the ILT: AI in environmental enforcement 

Our use case partner is the Dutch Inspectorate for Environment and Transport (ILT) , which is responsible for promoting clean air and water, including overseeing the transport sector on land and water. They're already exploring how AI can assist with their enforcement duties.

The inspectorate operates at the intersection of environmental protection, transport oversight, and regulatory enforcement. For them, AI is not introduced into an empty space. It enters an already complex governance environment shaped by institutional responsibilities, practical constraints, and diverse stakeholder interests. Understanding how AI functions within that environment is therefore as important as the technological capabilities themselves.

The potential is clear: AI could help detect violations earlier, predict risks before they materialise, and target enforcement more effectively. 

From polluters to partners: industry's dual role 

A massive socio-techno-legal system exists to keep our air and water safe. But for centuries, we've also allowed economic activity to degrade them. Today, industrial activity is the main focus of environmental regulation. Agriculture, transport, manufacturing - these sectors pose the highest risks to clean air and water. This puts industry in the position of being "sources of risk."

But industry can also be part of the solution in two ways: 

1. By producing technology: sensors and AI tools that detect environmental hazards 
2. By innovating processes: reducing pressure on air and water in the first place. 

Over recent decades, laws and regulations have increasingly placed precautionary responsibility on industry. The question is how to support that shift. 

Beyond the humdrum narrative 

Public debates about environmental harm often follow a familiar storyline: pollution occurs, authorities respond, and affected communities bear the consequences. While this narrative reflects real tensions, it can also obscure opportunities for earlier detection and shared responsibility.

A more proactive approach would emphasise collective capacity to recognise risks and coordinate responses before harm becomes irreversible. Industry actors, regulatory bodies, researchers, and civil society organisations each hold pieces of the information landscape necessary to identify emerging threats.

AI-supported sensing and monitoring practices may help connect these pieces, but technology alone cannot resolve the underlying governance challenges. Trust, transparency, and clarity about roles remain essential. In this sense, AI becomes part of a broader socio-technical system rather than a standalone solution.

The messy reality

An honest observation: we're far from this positive picture. Industry players often prioritise short-term profit. Authorities struggle with caseloads. Cooperation between NGOs and government agencies is hard to establish and maintain. When AI enters this space, it lands in that messy reality.

That's where ILT does its work. That's where our research happens. And that's why we need to understand not just the technology, but also the world it operates in. 

Opening the conversation

Exploring environmental harm as a dimension of public safety invites both conceptual and practical reflection. Conceptually, it challenges human-centred understandings of safety and highlights the importance of ecological conditions as foundational to public life. Practically, it raises questions about how emerging technologies can support authorities and other stakeholders in recognising and responding to environmental risks.

The AI MAPS use case on environmental harm aims to contribute to both conversations. We're examining how AI practices are currently used, how they are being developed and deployed, and how they might be anticipated in efforts to protect clean air and water in public spaces.

Because if public safety ultimately depends on the conditions that sustain life, then protecting those conditions is not a peripheral task. It is central to how we understand safety in the first place. 

This is the third use case in the AI MAPS ELSA Lab's research program on public safety. Previous use cases examined crowd control at demonstrations and feelings of safety in residential neighbourhoods.