Closed Circuit Television (CCTV) has been around for decades and has been a key technology for security agencies, police forces and transport providers.

Airports, shopping centers and city central business precincts today have hundreds, if not thousands, of CCTV cameras, but managing and using the technology has been reactive.

Finding shoplifters or analyzing accidents, for example, relies on reviewing past video footage, no matter how recent.

The advent of artificial intelligence (AI) has been a game-changer for video, and just as written data is transformed into AI applications through Large Language Models (LLMs), video providers are using Vision Language Models (VLMs), which treat images as data.

This enables solutions that can analyze hours of footage from a multitude of cameras – work which would previously have taken humans many hours.

Applications can also move into real-time with predictive applications.

Generative AI capabilities

One of the leaders in new video capabilities is Milestone Systems, founded in Denmark in 1998 and which has grown as video technology has advanced.

In April 2025, Milestone purchased the Berlin-based technology company brighter AI and is delivering new generative AI capabilities through Project Hafnia. This platform provides instant access to an ethically sourced, annotated video library, along with new developer tools.

Morten Illum, Milestone’s chief revenue officer, was in Australia recently, meeting clients and explaining how the new technology’s capabilities can be applied.

“What we are seeing is that there is a very big demand for moving into more proactive video,” said Illum.

“That goes across multiple verticals. From retail, in hospitality and in law enforcement, there is now an ability to use video to prevent things from happening or avoid things as they happen.”

Capturing video as data

AI’s capabilities begin with simplifying and shortening the time required to perform traditional tasks.

“If you operate a big airport and someone comes in the morning for their shift, and their job is to review what happened in the last six hours from hundreds of cameras,” said Illum.

“You can just say ‘tell me what I need to know, what I need to act on, what are the things that require janitorial assistance or need a security evaluation?’ Because we have captured the video as data, and this has been processed by the VLM, this is available immediately.”

“There is a very big demand for moving into more proactive video.” – Morten Illum, chief revenue officer, Milestone

There is also the ability for predictive prevention. The airport will have a protocol for how equipment should be positioned on the tarmac, including baggage handling and refueling equipment.

Using strict rule-based analysis, the video system can now detect when the equipment is not positioned correctly and when a dangerous situation or potential accident is about to occur.

Managing crowd dangers

This can also be applied to crowd management. If there are bottlenecks and potentially dangerous crowding in particular areas of a facility, the system can identify the danger before the problem escalates.

If a child is lost at the airport in real time, a brief description of the child, including keywords, can be entered into the system, enabling them to be found much faster than the traditional method, which has human operators scouring multiple screens.

A new development for the system is a Hafnia plug that automatically converts the video into short, descriptive text summaries of the action and can email them.

“If you need to send a description to somebody of what might have happened, instead of scrolling through videos and then writing that up, you can take a piece of video and send it to our engine, and it will deliver you a text summary,” said Illum.

Illum is at pains to point out that Milestone places a top priority on the ethical sourcing and use of video to ensure it doesn’t contravene privacy regulations.

Video is anonymized using capabilities from brighter AI, with the solution using generative AI to redact personal identifying information such as faces, license plates, and full bodies.

Illum says the journey is just beginning with the new capabilities, and Milestone’s open platform software enables customers and collaborators to develop their own applications.

“It’s a very exciting time in our industry,” he said.

“We are committed to staying open and delivering a platform that can help drive our industry forward.”

Image credit: iStockphoto/R_Type