What are the benefits, and drawbacks, of AI in security?

The integration of artificial intelligence (AI) in physical security is a critical conversation right now. Let’s start with the positives. Across the industry, AI is being used to automate routine tasks to give time back to security operators. This allows security operators to shift their perspective from being merely reactive to one that is proactive. This improves overall operational efficiency. Now for the drawbacks. While AI is allowing us to dream big, we’re often dreaming too big. AI is not going to replace the work of security operators. It should be used as an extra layer of protection, not the only layer. For AI to be effectively integrated into security, there needs to be clear communication about its capabilities and limitations, ensuring that its application in real-world contexts is practical.

Using artificial intelligence in physical security brings numerous benefits. Firstly, AI enables intelligent monitoring through features like smart surveillance cameras and facial recognition, enhancing surveillance efficiency and accuracy while reducing human errors. Secondly, AI can analyse big data in real-time, swiftly detecting abnormal behaviors, thus improving security alertness and response speed. Additionally, AI can perform pattern recognition, learning, and adapting to various environments, continuously enhancing the intelligence level of security systems. However, there are challenges and drawbacks to using AI. Firstly, the high cost may be a barrier to deploying AI security systems, including expenses for equipment procurement, maintenance, and personnel training. Secondly, privacy and data security issues require particular attention, especially concerning the collection and processing of personal information. Moreover, AI systems may encounter issues such as misidentification and missed detections, necessitating ongoing optimisation and algorithm adjustments to improve accuracy and stability.

The benefits of using AI have far outnumbered any potential drawbacks in our industry. The number of camera feeds in the world today is staggering. It’s simply not reasonable to expect humans to monitor them all effectively. AI helps organisations monitor these feeds 24/7/365 for anomalies and calls operators’ attention to events requiring further scrutiny, such as verifying an alarm. It also helps deter loitering and other undesirable behaviours by automating voice-down messages and turning on lights while escalating the response if the condition remains. The more we can automate routine processes, the more our customers can spend time on more important, customer-facing responsibilities. AI can also harvest valuable metadata for operations and sales departments to validate compliance or marketing campaigns. The main drawbacks can come from unscrupulous use of data that violates privacy or crosses ethical boundaries. It’s up to all of us to ensure AI is used responsibly.

AI technologies enable a range of powerful capabilities when it comes to physical security, such as detecting and classifying objects and their characteristics via video surveillance, as well as analysing audio and other activities in a scene. As AI continues to evolve, it’s able to improve upon the accuracy of analytics algorithms even further. For instance, with AI, these algorithms can detect objects even when they are partially hidden, or in challenging lighting conditions, providing an added layer of visibility and security. However, taking a risk-based approach to AI is critical, as there are several areas where the technology could pose risks to people and society if used improperly. These include data privacy (especially biometric data), potential bias, and automated processes or decisions that may be wrong. But by approaching AI in a responsible manner, it can be used to greatly enhance physical security systems and foster a safer world.

A benefit of using artificial intelligence in physical security is automation of certain tasks. This includes counting occupancy in a room or identifying if a camera’s view has been tilted or blocked by some obstruction such as leaves or a room re-arrangement. A drawback is the potential for over-reliance on the software. Artificial intelligence results should be reviewed by a person before a final decision to act is made.

Layering artificial intelligence (AI) over existing technology is going to reap huge benefits in the physical security space. In AV-over-IP technology, for example, we are looking at utilising AI to automatically create bookmarks when notable events happen in a video stream. This will alleviate the tasks of control room operators and add an additional layer of security. Of course, with new AI tools, there are drawbacks as well. It’s not crystal clear just yet how reliable AI tools can be with video footage. We could see an increased number of false alarms and more video footage being recorded than is necessary in control rooms. Overall, there are still a lot of unknowns about how AI is going to change the future of security. I foresee most of the results being positive ones.

There’s no doubt that AI has the potential to disrupt all kinds of industries. For physical security, AI is shifting systems from primarily being used as reactive tools for investigations into tools that can be used across a range of operational functions. Some of the exciting new use cases include warehouse and manufacturing settings. Before video security cameras were just used for investigating claims, but with advanced large vision models, they are now being used by cross-functional business partners to monitor for and send alerts when they detect abnormal or unsafe traffic patterns, missing safety gear (for example, someone not wearing a hard hat in a construction zone), and more. In retail settings, marketing teams are leveraging AI-powered motion and heatmap tools to understand the effectiveness of product displays and placements. With any new technology, we have to be thoughtful about risks and the rollout. We need more in-depth product education, as well as clear labels, controls, and permissions so that the teams that leverage these tools understand their limitations and expectations. 

AI is the force multiplier, making it possible for security teams of all sizes to shift to a proactive strategy. When preventing crime, detecting, deterring, and responding quickly is vital. Developers in the industry have reached multi-modality, combining computer vision, large language models, and other input signals to provide the AI with a near-human-like understanding of what’s happening in a video surveillance scene. This enables rapid response to threats, enhanced meta-data in the proactive stages, and conversational forensic searching. Some organisations are concerned about accuracy, specifically missed threat identifications or false alarms. The exciting news is that AI teams like ours tackle complex security application challenges daily. For example, parked vehicles have consistently challenged proactive video monitoring. Our team overcame this by developing new algorithms that nearly eliminate false alarms and focus only on moving vehicles. AI is paving the way for an exciting future for our industry.

By automating routine tasks and improving threat detection and response capabilities in cybersecurity, AI allows security teams to focus on more strategic activities. It is likely that cybersecurity professionals will soon use AI tools as an interface to less-familiar tools, similar to how programmers are currently using AI to generate code based on a task description when writing code for unfamiliar programming languages. However, one drawback of the rapid adoption of AI in enterprise security applications is that the accelerated development cycle and limited user base mean these tools have not seen extensive real-world security testing. As a result, some AI security tools may include vulnerabilities that can be exploited by threat actors. This is like the latest trends of enterprise file transfer applications or enterprise VPNs being compromised, which happened before the current increased development cycle to incorporate AI based on investor demands. It's very likely we will see zero-days used against AI-powered cybersecurity solutions in the short term.

Adept at recognising patterns and trends and recognising people, vehicles and objects, AI/ML can strengthen security by quickly and accurately detecting attempted intrusions, suspected criminal activity, and safety violations. The danger of using this revolutionary tool is the same as with using any other tool – using it inappropriately. Treating AI/ML as infallible or relying on it as the sole source of information, or – even worse – letting it make decisions for us, will lead to undesirable and potentially tragic results. Artificial intelligence is not taking over the world – not yet, anyway. But using this new tool incorrectly or not understanding that it is just a tool will lead to bad outcomes, for us and the technology. Misuse will result in public distrust and possibly banning of the technology, preventing the ability to leverage an extremely effective tool.

Artificial intelligence (AI) in corporate security offers tremendous efficiencies and enhanced capabilities, such as real-time threat detection and accurate anomaly identification, through technologies like anomaly detection systems (ADS) or license plate readers (LPRs). AI can also automate routine tasks, integrate with various security systems, and analyse large amounts of data to forecast potential risks, making operations more scalable and cost-effective. However, the technology is not without drawbacks. Key challenges include significant privacy concerns, dependency on quality data, high entry costs, and the complexity of system maintenance. Furthermore, AI can inadvertently create chaos if misused, for example, through AI-generated deep fakes or spoofed communications that can damage an organisation's reputation and economic well-being. While AI can significantly augment the capabilities of security practitioners by automating processes and sifting through data, it's crucial to emphasise that the critical analysis and proactive threat mitigation provided by human analysts remain irreplaceable.

Artificial intelligence offers robust opportunities for businesses to streamline day-to-day operations which allow employees to focus on big-picture thinking over more monotonous tasks. In regard to physical security, as an example, AI programs can consume CCTV video footage and discern known from unknown individuals, monitor human behavior patterns, and alert on anomalies; allowing physical security teams to monitor and respond appropriately to most important physical security threats. However, AI is far from perfect, and mistakes do happen. With AI utilisation in physical security, the stakes are much higher; often impacting humans directly. Similar to utilisation use cases of AI for generative or logical security purposes, Physical Security AI also needs to have close human oversight with defined responsibility for model tuning, fairness, and validating alerts or outputs. These safeguards not only help ensure bias free detections; they also help ensure civil liberties and human rights are not infringed upon. 

The benefits of using AI have often been stated, including significantly reduced false positives with analytics thanks to accurate AI-based object recognition of people and vehicles. As technology has evolved, the ability to recognise custom objects has provided even more ways for organisations to track objects reliably. AI has enabled security and operations departments to become more proactive to threats in real-time versus merely reacting to events that happened in the past. By evolving security cameras into sophisticated IoT devices, AI has become a capable “assistant” watching over multiple video streams 24/7. Performing such a mundane task allows humans to spend time on more important responsibilities while ensuring important events are not missed. Any drawbacks would be due to improper use of AI technology, such as collecting and analysing large amounts of personal data, leading to privacy violations. Improperly trained AI systems can inherit biases in their training data, leading to discriminatory practices.

Artificial intelligence is transforming the industry, making this an exciting time for businesses. AI offers numerous benefits, such as enhanced business intelligence for security, alarm, and fire installers, and supports the future of these businesses with AI-driven services. Dealers can use data from intelligent solutions to manage accounts more effectively and verify alarms, reducing false positives. AI also creates new revenue streams through high-value services like remote and proactive video monitoring. However, we are just beginning to explore AI's potential. The most valuable use cases will be determined through testing and time, revealing the true impact of this technology.

Innovations in technology are transforming physical access control, significantly boosting security and efficiency. These systems can proactively identify breaches or anomalies, automate reporting, and analyse data to detect access trends. For example, determining the busiest check-in times in a building can be done seamlessly, optimising resource allocation and enhancing security measures. However, we must consider the larger implications of AI. For example, it's important to address the need to protect personal data and maintain the security of any systems. As technology continues to advance rapidly, finding the right balance between the benefits and potential drawbacks of these technologies will be crucial for the most impactful implementation.

The growth of AI continues to drive significant advances in biometric identification, especially contactless finger and face recognition. Data shows that the technology attracting the most significant research and development spend is facial recognition across several forms including mobile, CCTV, passport/license matching, and dedicated access control. The uptake of contactless biometric technologies (particularly facial) has been heavily influenced by COVID-19, due to sanitation requirements and the need for identity verification whilst wearing a mask. While AI offers many benefits in terms of advanced access permissions, ethical and cost considerations do exist. Ethical use of facial recognition for crime and stadium use is under constant review globally, with varying degrees of adoption among different countries. Obtaining Privacy Impact Statements from customers prior to their business adopting biometric technology is encouraged, and customers should expect to pay more for their hardware to achieve the benefits of AI.

Artificial intelligence (AI) has become increasingly prevalent in physical security systems, offering various benefits but also presenting some drawbacks. A key benefit is enhanced threat detection, where AI can analyse large amounts of data and accurately detect potential security threats much quicker than human operators. This leads to improved response times and reduces false alarms. The drawbacks include privacy concerns (as AI gathers sensitive data), potential bias and discrimination, and the high initial costs of deploying a latest generation AI solution.

There is little doubt that AI is having a massive impact on physical security as a whole, in nearly every aspect of the market. It’s becoming a major selling point and an in-demand feature, even though a product can use AI in a vast variety of ways. For example, the use of AI can be overt – maximising efficiency in data analysis – or used in the background to improve the quality of the video based on the scene analysis. Like other technologies such as cloud, AI presents a vast array of opportunities, and how and when you choose to implement these can make a big difference in your business. For now, it’s both a buzzword and a reality, and we need to educate the market of the wide array of AI applications and the actual impact on current and future technology.