Building Safer Spaces: How AI Is Changing Threat Detection with Steve Roskowski

As public safety threats escalate across schools, workplaces, retail spaces, and community areas, one innovation is gaining significant momentum: AI-powered gun detection.  It has quickly become one of the most critical and widely discussed technologies of our time. Its momentum is driven by a clear and urgent need: detecting threats early enough to prevent them from escalating into tragedies.

AI gun detection transforms standard cameras into intelligent systems that monitor continuously, recognize weapons in real time, and alert security teams within seconds. This not only strengthens preparedness but also helps humans respond faster and more effectively when every second counts.

The pace of advancement in this field has accelerated rapidly. Breakthroughs in edge computing, cloud AI, synthetic data generation, and advanced model architectures have dramatically improved both accuracy and response speed. As a result, real-time gun detection is no longer a theoretical capability – it is scalable, reliable, and already being deployed across high-risk environments.

And with the majority of Americans supporting AI-powered gun detection in public environments, the demand for ethical, transparent, and highly reliable solutions is higher than ever.

At the forefront of this transformation is Steve Roskowski, Chief Technology Officer at Eagle Eye Networks. With decades of experience in connected systems and deep expertise across software and hardware systems, Roskowski brings one of the most authoritative perspectives on AI-enhanced security.

AI Gun Detection Technology

With strong public support for deploying gun detection in schools and workplaces, how is Eagle Eye Networks using edge AI and cloud technology to deliver real-time, reliable gun detection?

Roskowski: Yes, 77.4% of Americans support the deployment of AI-powered gun detection in schools, workplaces, and houses of worship.

There are millions of security cameras in the world today, but they are not generally used for proactive security. Eagle Eye Gun Detection can transform existing security cameras from passive recording devices into a proactive gun detection system. To be effective gun detection needs to be very fast and extremely accurate. We engineered Eagle Eye’s Triple-layer Verification system to achieve the highest levels of accuracy.

This is how it works:

Rapid Edge AI:

The Eagle Eye AI Bridge analyzes video from security cameras to spot potential brandished guns. It is located at the school or business. It uses an AI model appropriately sized for widespread edge deployment and instant identification of potential guns. We specifically error here on “permissive” so we do not miss events.

Advanced Cloud AI:

When a potential gun is detected, appropriate video is instantly transmitted to the Eagle Eye Cloud. A large-scale Eagle Eye AI model analyzes it to ensure accuracy and reduce false positives.

Eagle Eye Human Review Service:

When the Advanced Cloud AI detects an actual gun, the Eagle Eye Human Review Service, which is staffed with trained professionals, verifies that a gun is present. As soon as that happens, the customer’s emergency response plan is executed. The response typically includes SMS messages, phone calls, and
Emails. These are sent to key security or company personnel, law enforcement, response service, key executives, and emergency responders.

According to you, what recent advancements in AI have most improved the accuracy
and the speed of gun detection systems?

Roskowski: Advancements in AI, both on the edge and in the cloud, are happening faster than anything we’ve seen in any industry in history, and that is only the tip of the iceberg.

You have Moore’s Law at the very bottom driving massive improvement in hardware performance for running models, dramatic improvement in the model design and architecture, and fundamental innovation for things like LLM, all with hundreds of billions of dollars in investment driving it forward. So, there’s little wonder that accurate gun detection is becoming viable.

Meanwhile, recent architecture improvements on Convolutional Neural Networks (CNNs) and the subsequent development of Large Language Models (LLMs) and Vision Language Models (VLMs) are profoundly impacting all aspects of AI Ops, including data curation and validation.

The ability to create synthetic data to augment real-world data dramatically increases the quantity and quality of training data, leading to fast development of specialized models that deliver great accuracy.

Combined with the hardware and model architecture improvements, we can now run very accurate models on low-cost edge hardware. We can also deploy very accurate and efficient models on cloud infrastructure to accomplish our second-tier validation. None of this was doable two years ago.

How do you ensure that AI-driven gun detection remains effective across diverse
lighting, camera angles, and crowded environments?

Roskowski: AI and automated detection watches constantly, so the system can be much more “alert” than human monitoring. With Triple-Layer Verification we can also tune the edge AI to be very sensitive.

This means we get a lot of first-layer “maybe” results, but still enough filtering to allow a more accurate (and compute intensive) second layer to run effectively. Even there, we incorporate a final layer–the human in the loop—to avoid false detection. No single layer can adequately ensure effective detection. The total system design addresses the challenges inherent in crowded environments, and elements such as lighting and camera angles.

That said, setup and configuration are still key levers to increase accuracy. Importantly, Eagle Eye Networks always sells through our global network of professional security integrators who ensure that our systems are properly designed and installed.

Our security integrator partners have extensive experience working with security cameras in a variety of environments and know how to optimize camera angles and reduce problems with lighting. We believe professional installation is essential.

Ethics, Privacy and Public Safety

When it comes to responsible AI usage – what safeguards and governance principles guide your product development?

Roskowski: At Eagle Eye Networks our mission is to deliver cybersecure, cloud-based, AI-powered video surveillance to make businesses more efficient and the world a safer place.
We are committed to the ethical and secure use of AI to improve security, safety and commercial efficiency. Our R&D, products, and processes align with that mission. And we believe products such as Eagle Eye Gun Detection can help make our communities safer, potentially save lives, and help address some of society’s most persistent problems.

However, we are also keenly aware of the possible misuse of AI technology and other technologies. We believe it’s important to be transparent and forthcoming about how AI technology works and is used, and to build cybersecurity and privacy protections into products and processes.

Before founding Eagle Eye Networks, our CEO Dean Drako founded and served as CEO of cybersecurity company Barracuda Networks. As a result, cybersecurity and privacy are part of our DNA. A strong encryption and privacy protection model is built into our system at the base layer. All media is fully encrypted in transit and at rest to help prevent any misuse of the content. We use a broad range of datasets to avoid bias in the models.

Another aspect of our gun detection’s Triple-Layer Verification is that the multiple perspectives (two layers of AI and human verification) alleviate bias that could emerge in the AI models. In general, our approach is to balance the level of “observation” with the problem being solved.

Public safety in high-risk areas like schools and malls justifies the additional scrutiny of the full AI-enabled monitoring we use for gun detection. Every second counts in an emergency, and proactive systems such as Eagle Eye Gun Detection can help mitigate or possibly avert tragedies.

How can AI gun detection enhance public safety without crossing into over-surveillance, and how important are transparency and public trust in its adoption?

Roskowski: No single technology can solve the problem of gun violence at our schools, houses of worship, and workplaces. Gun detection is a key part of a layered approach to security which includes visual gun detection, door access control, metal detectors, panic buttons, training of staff and students, execution of practice drills, fencing, secure areas, other physical barriers, and human oversight.

Transparency and being forthcoming about technology capabilities are essential in the successful deployment of any technology and are key to engendering public trust.

Cybersecurity and privacy protection should be inherent in product development. But they also need to be part of the deployment of technology, and the internal processes of the schools, retail spaces, and cities as well. This is another reason why professional installation of physical security systems is important. The professional installer works with the school or business to ensure that the system aligns with local and institutional regulatory and compliance standards.

Importantly, a cloud video surveillance system can actually help organizations reduce liability, adhere to local laws and regulations, or meet insurance requirements.

Innovation, Leadership and Vision

How do you see the CTO’s role evolving in an era where AI innovation is moving
faster than regulation and public readiness?

Roskowski:  My technical career has spanned the greatest technical innovations ever in history. I’ve witnessed the move from wall phones and black-and-white TVs to the internet and ubiquitous smartphones, and now AI.

The role of a CTO is to understand what technology is capable of and how a company should be reacting to that. What’s changed for the CTO isn’t the role, it’s the pace. What previously was a five-year projection is now a five-month projection.

In addition, the technology is changing much faster than regulations can be promulgated to govern its use. This isn’t the first time this has happened; again, it’s the pace of change that’s new. AI is advancing at a meteoric pace. This fact raises the importance of ethical and secure R&D, transparency, and proper deployment of technology.

At Eagle Eye Networks, we support the work of the Security Industry Association (SIA), which champions the development and use of AI in ways that are “human-centric, ethical and trustworthy and that mitigate potential risks”. SIA recently issued guidelines for the use of AI in physical security.

What’s your personal approach to fostering creativity and invention inside a fast-
growing tech company?

Roskowski:  Focus on personal empowerment, build confidence, and encourage the ability in each person to be inquisitive and find joy and professional fulfillment. Engineers are fundamentally artists. Instead of painting, our path to creativity and invention is writing code and building things. Trust and belief in yourself will drive everything. Interestingly, AI empowers the work of an engineer–it can make invention a lot less time consuming. You can paint faster since you have a magic brush, but it is still your vision and ideas driving it.

Future perspectives

If you could design the ‘ideal future’ of intelligent security systems 10 years from
now, what would it look like?

Roskowski: Enhanced with AI, security cameras–even existing cameras–can detect a fire, somebody falling down or someone doing something unusual. You’ll still need door locks, temperature sensors, gate sensors and other equipment. But the security system a decade from now will not be a set of siloed data sources.

The intelligent security system of the future incorporates and analyzes different data and event sources in one system that proactively looks for problems (fire, someone falling down), anomalies (a person in a place they don’t belong or after hours), and situations of interest (more people shopping for cars in the morning). It keeps an eye on things for you. Whether you’re a business owner or school administrator, it immediately brings issues to your attention.

The system provides full context of the situation. For example, “This door opened, two people entered at the same time, and only one is an employee.” The system can also take automatic action to notify people, send emails, contact law enforcement or janitorial services, etc.

It’s security that’s proactive, an AI Safety and Security System. The first iteration is available today with products like Eagle Eye Gun Detection.

What advice would you like to give to future tech leaders?

Roskowski: Welcome, and get on board the AI train. Technology is evolving so fast; if you are not aggressively adopting AI, you’re behind. The pace of change that we’ve seen in the last 50 years is going to feel like molasses compared to the pace of change we’re going to see over the next 50 years.

Closing thoughts

Integrating AI gun detection technology isn’t just another security upgrade – it represents a meaningful shift in how organizations identify and respond to threats. By integrating real-time visual intelligence into existing camera systems, this technology enhances both awareness and response capabilities at critical moments.

For leaders, adopting AI tools like gun detection should be part of a larger strategy to assess the entire security technology stack. They need to understand where vulnerabilities exist, how systems connect, and which solutions offer the best balance of value, reliability, and risk.

Gun detection alone isn’t the complete answer

But when used thoughtfully, it becomes a high-impact component of a modern, data-informed security ecosystem.

However, genuine progress requires meaningful collaboration among security/tech leaders, policymakers, educators, and local communities. It requires not just better tools, but better alignment on privacy, ethics, and the role AI should play in protecting public spaces.

As Steve Roskowski emphasizes, intelligent safety is no longer a future concept — it’s already here. But its impact depends entirely on how wisely and responsibly today’s decision-makers implement it.

The stakes are high, but the opportunity is greater. Leaders who act today will define what safer environments look like for the next decade – and ensure that life-saving technologies reach the people and places that need them most.