printer friendly version

AI means proactive surveillance

Artificial Intelligence (AI) has transformed video surveillance systems, significantly enhancing their efficiency and accuracy, while enabling proactive threat detection. A fundamental aspect of AI in surveillance is its capability to automatically analyse video footage in real-time and identify events of interest. Although this technology is still in its early stages, it is continuously advancing.

AI also offers valuable data-driven insights that support various operational needs, marketing strategies, and other non-security functions within a business. This includes analysing foot traffic patterns, customer counts, heat maps, peak wait times, and parking occupancy. Moreover, some well-managed cities are even adopting AI-driven solutions for traffic management, aiming to optimise road usage and enhance public safety.

Furthermore, AI systems enhance human efforts in control rooms by operating around the clock and identifying alerts that may otherwise go unnoticed. While some alerts may be false positives, advances have been made to effectively address this challenge.

AI solutions are offered for any installation architecture, on-premises, cloud, edge and hybrid. However, due to the ever-changing nature of the technology, some claim that cloud-based AI is a better option as updates are automatically applied without requiring any action from the user, ensuring you automatically use the latest technology. SMART Security Solutions asked DeepAlert, an established AI-in-the-cloud provider, for some insights into AI and its current and future applications.

How has AI transformed video surveillance and analytics?

DeepAlert has witnessed and driven the evolution of surveillance from passive monitoring to proactive, intelligent systems, says the company’s, Mark Smuts. “AI has radically transformed video surveillance by enabling real-time analysis at scale, reducing false positives, and empowering operators to focus on genuine threats. Today, monitoring thousands of cameras is not only feasible, but efficient – thanks to AI.”

In the next 12 to 24 months, DeepAlert expects to see a shift from rule-based systems to autonomous, agentic AI, capable of deeper scene understanding and multi-camera contextualisation. These systems will not just detect anomalies, but understand environments dynamically, responding appropriately across multiple sites.

While behavioural analytics is still complex and resource-intensive, it is already making a measurable impact in high-value sectors such as retail loss prevention. As technology matures, behavioural detection will become more accessible, helping pre-empt security incidents before they escalate.

Edge, onsite or cloud?

The decision on whether AI is located in the cloud, run by third-party service providers, on edge devices, or within on-premises solutions is up to the policies of the end user, as well as their privacy and security requirements. The reality is that hybrid solutions are the most likely solution, something that most cloud users have discovered in the last year or two where some companies (especially in the US and EU) threw their whole technology infrastructure into the cloud, only to discover that may not be the optimal or most cost-effective solution.

Smuts supports this view, noting that AI is being deployed across edge, server, and cloud environments, each suited to specific use cases. He says that edge devices are optimal for fast, localised detection tasks (e.g., intrusion), while cloud platforms offer scalability, flexibility, and the ability to aggregate and analyse data from multiple sources and locations.

“The future lies in cloud-first or hybrid models, where complex tasks like behaviour recognition or cross-site scene analysis benefit from the processing power and accessibility of the cloud. Ultimately, customer needs and security requirements will dictate the architecture, but cloud-based AI is emerging as the optimal platform for advanced analytics.”

Who makes the final decision?

Being exposed to science fiction movies and a lot of hype around AI, it is not unreasonable to expect that AI can be left to its own devices to make decisions without human input. Of course, anyone who has asked one of the public AI services a question soon discovers that, while they are remarkable and have improved dramatically, there is a reason they all have a footnote warning that they may not provide 100% factually correct answers. So, in the midst of these impressive developments, how far can we go in ‘trusting’ AI?

“While AI has become highly reliable, complete autonomy in decision-making still faces hurdles around trust, liability, and explainability,” states Smuts. “At DeepAlert, we believe in agent-assisted automation, allowing AI to prioritise alerts and actions, while humans retain oversight over critical decisions.

“In time, as integrations between surveillance, access control, and dispatch systems improve, AI agents will become increasingly capable of managing security operations end-to-end – from early detection to initiating emergency response protocols, such as locking down a facility or notifying emergency services.”

This is not to say AI does not assist and optimise a surveillance operation. It is pushing security operations from a reactive posture to a predictive, preventative model. By analysing historical patterns, real-time data, and environmental context, AI systems can identify vulnerabilities and signal potential threats before they materialise.

For example, Smuts explains that AI can monitor camera health and notify operators if a field of view is obstructed or tampered with, a small, but critical step in ensuring operational continuity. “In more advanced applications, AI can detect suspicious behaviour patterns across time and space, enabling early interventions and crime prevention strategies.”

The privacy question

The question of privacy is always top of mind when discussing AI. While privacy is a critical question, the world has become used to operating system and social media companies sucking up private and sensitive information without any restrictions (read the terms and conditions if you disagree). Of course, other companies do the same and our internet usage and habits are broadly shared across the world, despite some regulations.

That said, AI can invade one’s privacy better than older technology and the need for ethical usage is important – although, when one looks at how companies in general treat users’ data one has to ask if the concept of ‘ethical’ is even considered beyond the marketing material? Some companies have implemented AI privacy algorithms to enable effective surveillance, while protecting privacy (see the Milestone announcement here http://www.securitysa.com/24616r.).

Far from undermining privacy, Smuts says that AI can enhance it when implemented responsibly. “Technologies such as face blurring, anonymised detection, and privacy-by-design architectures are already in use and becoming industry standards. As global regulations evolve, the ability to deliver security, while upholding privacy will become a key differentiator. DeepAlert embraces this direction, ensuring our systems are built with trust, transparency, and data protection at the core.”

Redefining the future of surveillance

When looking at what DeepAlert is doing with its AI services, Smuts says, “at DeepAlert, we are redefining the future of video surveillance through innovation and collaboration. Our AI platform, already trusted for intrusion detection, now supports advanced models.”

Some of the capabilities currently supported by DeepAlert include:

• Open gate detection.

• Fire and leak alerts.

• Prone person detection.

• Virtual guard tours.

• Automated site compliance checks.

“These capabilities are designed to reduce operator workload, boost response times, and make security smarter and more proactive,” concludes Smuts. “With the integration of next-gen AI and GenAI technologies, we are solving complex monitoring challenges with accuracy, speed, and cost-efficiency.

“We collaborate closely with leading security operation centres and private security firms to tailor our solutions to real-world needs, ensuring our technology not only detects threats, but empowers operators with actionable intelligence.”

Share this article:

Further reading: