The Fast Pace of AI Innovation

From Vision Language Models to Agentic AI, discover how the next generation of multimodal systems is transforming threat detection and response.

• By David Antar
• Mar 06, 2026

The pace of innovation in AI-enabled security technology is accelerating faster than the speed of light. Where major breakthroughs in artificial intelligence once occurred every few years, they are now arriving every quarter — and in some cases, even faster.

This rapid evolution presents both an opportunity and a challenge for security professionals to navigate the future of AI and security. On one hand, AI is unlocking powerful new capabilities in threat detection, situational awareness and automated response.

The Volume of New Technologies

However, the sheer volume of new technologies can make it increasingly difficult for organizations to understand what is real, what is hype, and where to invest.

No matter what industry you work in, individuals face a plethora of threats as soon as they step outside their front door, as security threats have become part of our normal everyday existence. With different threats facing each industry, one common denominator is detecting the threat before it becomes an actual incident.

Over the last few decades, the security industry has seen significant growth when it comes to video security and predictive analysis. From the first IP camera launched in 1996 by Axis, which, ironically, we were the first U.S. dealer, to the current levels of rapidly evolving AI-enabled video analytics, it has been a firestorm of innovation.

With the next generation of AI and behavioral analytics, we are watching the start of a new evolution, helping organizations make better predictions on the likelihood of future events taking place.

With most organizations dealing with budget constraints and resource challenges, AI, properly deployed, is helping them do more with less, as AI has become the eyes and ears for organizations rather than a physical person. With minimal human intervention, AI can help detect a threat by using data analysis to help organizations make intelligent decisions.

Emerging AI Technologies

Although researchers have worked on AI for decades, it burst into public consciousness with the launch of ChatGPT in late 2022. Within two months, it reached 100 million users — faster than Instagram or TikTok. Today, more than one-third of U.S. adults use generative AI tools, and adoption is even higher among younger professionals.

In the security industry, the next wave of innovation is being driven by two closely related technologies: Vision Language Models (VLMs) and Agentic AI.

Multimodal AI Systems

Vision Language Models are a new class of multimodal AI systems capable of processing video, images and text simultaneously. The term multimodal refers to the ability to process more than one “mode” of input and output (e.g., video and text). Unlike traditional video analytics, which can only identify objects such as a person, vehicle or weapon, VLMs can interpret a scene and describe it in natural language.

For example, rather than simply flagging “person with object,” a VLM can generate a real-time description such as “A man dressed in black is walking across the parking lot, carrying a rifle toward the main entrance.”

Agentic AI is also reshaping the landscape of video security today. Traditional AI systems focus on detection: identifying motion, recognizing objects and generating alerts. Agentic AI goes several steps further. It can see an environment, interpret behavior, make decisions, take actions, learn from outcomes, and coordinate with other AI agents.

Agentic AI can automatically lock or unlock doors, dispatch guards, trigger public address systems, initiate emergency protocols, and coordinate responses.

Humans-in-the-Loop

One of the biggest challenges facing users of AI-enabled security technologies is distinguishing between hype and reality. While many new products look as though they have been pulled straight from a science fiction movie, there is still a huge discrepancy between technology that looks impressive at a tradeshow and that which is best in a real-world scenario.

For example, a video security system that “most of the time” detects active shooters is not very reassuring to school administrators, teachers and parents. Conversely, a system that consistently flags cellphones and purses as handguns will lead to false alarms and ultimately be ignored.

Once we filter out the marketing hype, the reality is that even the most powerful AI systems are not 100% dependable, all the time in all conditions. This is why there is still a place for highly trained humans to verify potential safety and security threats and escalate appropriately: whether to “talk down” to an intruder or call first responders.

The AI-industry term for human verification and validation is Humans-in-the-Loop (HITL). Industry’s best practice is to use HITL both in real-time (during incidents) and offline to review image data used to train the next generation of AI models.

“Just like newborn babies, AI needs training to understand and interpret the data correctly,” said Charlie Fleischman, general manager at ViewScan. For example, AI can interpret that something is not right in a situation and flag it as amiss.

It could be too many people running in a certain direction, a person falling to the ground or a person holding a gun that can trigger AI to send alerts about a possible risk. However, for AI to recognize an issue accurately, there needs to be significant human training on exactly what to look out for daily. The goal is to minimize false positives and false negatives by continually training the model.

A Smarter, Safer World

With the evolution of AI-based systems from “object detection” to “behavior detection” and now predictive analytics, what is next on the horizon? We are seeing AI expand well beyond video, with integrations into building management systems and other third-party platforms.

To help bring this vision to light, ViewScan was formed through longstanding partnerships among experienced technological professionals, law enforcement and security guard firms. ViewScan’s software platform provides constant security and serves a wide variety of markets, including K-12 schools, higher education, healthcare, municipalities, HOAs, gated communities, car dealerships, commercial buildings and many more.

“At ViewScan, we believe that artificial intelligence has an incredible role to play in augmenting and scaling – but not replacing – the capabilities of security professionals and first responders,” said Jonathan Klinger, strategic advisor at ViewScan. “At the same time, we believe that highly trained remote security guards are needed to deliver the highest level of safety and security. This is why we have invested in a virtual security operations center staffed by TMA-Certified disabled American veterans and active-duty spouses. We also use a HITL workflow for continuous improvement and to ensure the accuracy of our models.”

ViewScan continues to be at the forefront of AI technology and is making significant strides to better protect people by building an open ecosystem to take advantage of the latest advances in AI.

Safety is enhanced through human presence, object, and vehicle detection, real-time, proactive remote monitoring, delivery service screening, guard booth operations, and construction site monitoring—delivered via fixed installations, mobile security trailers and enclosure-mounted solutions.

A Pivotal Moment

This is an extremely exciting time for the security industry. Near-term developments such as Vision Language Models, Generative and Agentic AI will revolutionize our ability to protect customers, employees and communities. The security industry is entering its most significant transformation since the move from analog to IP video cameras. Over the next two years, solutions using Vision Language Models, Generative AI, and Agentic AI will fundamentally reshape how organizations protect their people, property, and communities — enabling higher levels of situational awareness and faster response without proportional increases in headcount or cost.

Companies need to make smart investments in technology that continuously and dynamically improve to deliver the best return on investment (ROI). The possibilities of AI are endless and will continue to pave the way for organizations to mitigate risks by enabling smarter decisions with the right data in front of them.

This article originally appeared in the March/April 2026 issue of Security Today.