AI and Machine Knowing: A new challenge for CPTED and crime prevention

This Saturday's ICA Masterclass on CPTED White Papers will explore frontier AI systems, machine learning, and the pros and cons of AI and CPTED. The session is a primer for the AI and CPTED sessions at the 2026 CPTED conference, Oct 2-4 in Calgary. 

by Gregory Saville

Artificial intelligence is arriving in crime prevention faster than many practitioners realize. We’ve been posting blogs about AI in crime prevention for years. In my 2023 blog, Stop Dave, I’m Afraid, I reported on Smart City research by Finnish AI researchers.  

Last year, in Gambling with the Future I summarized Nobel Laureate, Professor Geoffrey Hintons’ warning at the latest AI conference. In What 1980s Weather models taught me about crime prediction I described my own research experience with scientific weather forecasting models. 

Today the story has expanded again. From predictive analytics and digital twins to large language models and autonomous surveillance systems, the question is no longer whether AI will influence CPTED. The question is how.

For decades, CPTED practitioners have focused on the relationship between people and places. We have examined territoriality, access control, surveillance, maintenance, social cohesion, and neighborhood capacity. Yet frontier AI systems introduce something new into that equation: machine knowing.

ARE YOU USING A FRONTIER SYSTEM?

Frontier systems include leading generative LLMs such as ChatGPT, Claude, and Gemini, along with advanced multimodal systems that work with images, audio, and video. They are less common in predictive analytics, such as hotspot mapping and crime forecasting. Some frontier systems are now becoming agentic, meaning they can pursue goals with limited supervision, although these remain in the early stages of deployment.

Frontier systems raise a new question: What happens when a machine begins interpreting places, risks, and human behavior alongside, or instead of, human observers?

That question forms the foundation of the Masterclass this Saturday. The session serves as a primer for the larger AI and CPTED discussions planned for the ICA Conference in Calgary this October.

AI-assisted auditing may help practitioners identify environmental risks, but human interpretation remains essential - AI-generated imagery for this blog 

SATURDAY’S MASTERCLASS 

On Saturday, the Masterclass begins with an unusual exercise. I begin with a conversation between myself and the same AI system chatbot I used throughout the development of the CPTED and AI White Paper. The purpose is not to demonstrate technology, but to explore a deeper question. How should we relate to an intelligence that can reason, explain, persuade, and create knowledge while making mistakes that appear convincing?

One of the most important concepts discussed will be the phenomenon known as a confabulation rate. In 2023, I described the problem of AI hallucinations. Confabulation takes that concern a step further. 

It refers to the frequency with which an AI system generates information that may be partly true and sounds plausible but is incorrect or entirely fabricated.

The future of prevention may depend less on what AI knows and more on how communities use that knowledge - AI-generated image for this blog

WHY DOES THIS MATTER IN PREVENTION? 

Confabulation matters because prevention professionals increasingly rely on digital information. If AI is used to analyze crime patterns, evaluate risk, generate recommendations, support neighborhood planning, or advise decision-makers, practitioners must understand both its capabilities and its limitations. Errors like confabulation rates may never vanish, so we must understand when and how they occur.

The opportunities and risks are extraordinary. The future of CPTED and crime prevention will not be shaped only by better environmental design or social cohesion. It will also be shaped by how we choose to work with increasingly powerful forms of machine intelligence.

Join me on Saturday, June 27 as we begin exploring that future. 

Gambling with the future

The shape of digital thinking - not always what it seems - Stockphoto 

Techno-solutionism 

(noun) tech·no·so·lu·tion·ism |ˈtek-nō-sə-ˈlü-shə-ˌni-zəm *
1: The idea that every social problem has a technical fix — a mindset that often bypasses community input, local culture, and lived experience in favor of data-driven shortcuts. 

by Gregory Saville

Do you know what’s happening in Las Vegas next month?

More gambling and fantasy — except this time, it’s not about blackjack or slot machines. It’s about our future.

The Ai4 Conference is next month — one of the biggest gatherings of AI developers, investors, and corporate clients in the world. It’s packed with major institutional players: banks, financial giants, data brokers, and tech developers. They’re not there to sound alarms. They’re there to map the future — one algorithm at a time.

But not everyone is buying in.

One of the voices at this year’s event is Geoffrey Hinton, widely regarded as the godfather of deep learning. He’s a recipient of the 2024 Nobel Prize in Physics, and he resigned from Google in May 2023 — not in protest of bad engineering, but in alarm over what his own research might unleash.

He’s not there to sell. He’s there to warn.

And if even Hinton is nervous about the direction of AI — especially when it comes to ethics, autonomy, and control — we should all be paying attention.

That includes those of us working in urban safety and crime prevention.

Geoffrey Hinton at his Nobel Prize ceremony
Photo by Jennifer 8. Lee, CC BY-SA 4.0, via Wikimedia Commons

AI AND CPTED - THE SILENT CONVERGENCE

Artificial Intelligence is already reshaping how cities think about safety. I described this convergence in detail during my keynote address on AI and CPTED at the 2021 CPTED Conference in Sweden. It’s happening in ways both subtle and systemic: facial recognition networks quietly expanding in public places; real-time surveillance feeding data into centralized dashboards; algorithms determining who is suspicious and who is "safe".

We’ve raised alarms about these trends before on the SafeGrowth blog — in posts like:

• What 1980s Weather Models Taught Me About AI 
• Summoning the Demon: AI in Law Enforcement 
• The Pros and Cons of Using AI to Prevent Crime 

My keynote at the 2021 conference was titled: Artificial Intelligence, Smart Cities, and CPTED. A Threat to the ICA.

Keynote address at the 2021 ICA CPTED Conference - Sweden 

Much of this unfolds behind the scenes. And yet, it touches our daily lives, our neighborhoods, and the very public spaces that CPTED  seeks to protect.

The problem? These AI tools are often built through a corporate techno-optimist lens. One that:

• Prioritizes efficiency over ethics
• Amplifies surveillance capitalism
• Ignores spatial injustice
• Replicates systemic bias in code causing algorithmic harm

If we’re not careful, we risk building cities that feel more like open-air data farms than thriving communities.

Marriotte Hotel at the Las Vegas convention Center - location of the 2025 AI conference - photo by Marriotte

THE RESPONSE? ETHICS BEFORE ALGORITHMS

That’s why the International CPTED Association (ICA) has convened a new AI and CPTED Subcommittee — a global collaboration of criminologists, computer scientists, planners, ethicists, and practitioners (myself included) working to confront these challenges head-on.

I’m chairing this subcommittee, and after months of research and reflection, the ethical questions have crept into my consciousness like invasive roots spreading beneath the roadways of our communities. You don’t see them right away. But they’re there — widening the cracks.

We’re now finalizing a white paper to guide the CPTED field — a direction-setting document for how to responsibly navigate this new terrain.

RECLAIMING SAFETY IN THE AGE OF AI

This isn’t just a matter for data scientists or tech firms. There is a message shaping up for CPTED practitioners and residents alike. To me, that message goes like this:

We must call for oversight, ethical scrutiny, and community-driven approaches over digital or predictive models that falsely claim certainty without embracing uncertainty and probability. 

And above all - techno-solutionism must not prevail. Cities are not math problems to be solved by code. They are ecosystems of people, stories, and space — deserving of care, not automation. Despite what some experts claim, sustainable crime prevention is not a simple matter of quick situational fixes. Livability too is important.

We’ll be sharing more soon in our ICA white paper.

For now, the dice may be rolling in Vegas. But the rest of us?
We’re not gambling with our future.