New York City Mayor Eric Adams recently said on Good Morning America that he wants to install artificial intelligence-driven weapons detectors in the subway system. Evolv Technology said its AI scanner could replace conventional metal detectors, bypassing the need for people to stop and empty their pockets. “With AI, we are able to detect many different objects of all sizes and types. Detecting if someone is carrying a gun might be highly beneficial to prevent attacks as well as to detect threatening situations,” Mikaela Pisani, the chief data scientist of AI company Rootstrap, told Lifewire in an email interview. “This would give authorities more time to take action before the situation escalates further.”

AI That Checks You Out

Evolv’s technology uses radar and lidar light-emission techniques to create images that an AI then examines. The company says its system can identify a concealed weapon on someone who walks through the scanner and prompt security to intervene. The technology “spots concealed weapons and other threats using advanced digital sensors and artificial intelligence,” the company writes on its website. “It’s incredibly accurate and can screen up to 3,600 people per hour—10 times faster than traditional metal detectors.” Pisani said that the most common approach to identifying people with guns is the use of deep learning models using CNNs (convolutional neural networks). CNN’s are designed to identify images by processing pixel data. Machine learning models are trained on large datasets of images, with different situations in whether people are carrying guns or not.  “With this method, classification systems are built to detect situations where guns are being used,” Pisani added. “More specific models might be used like object segmentation. to identify in which part of the image the gun is located.” Evolv doesn’t provide much public information about exactly how its system works. But Stephanie McReynolds, head of marketing at Ambient.ai, a computer vision intelligence company that automates physical security operations told Lifewire via email that similar systems analyze live surveillance video streams, using AI in conjunction with computer vision algorithms to detect objects like a person or a firearm. The computer looks at interactions between those objects which create signatures of physical movement or behavior.  “This deep analysis of threat signatures enabled by computer vision intelligence can provide rich context that goes beyond location or the presence of a firearm,” McReynolds added. “Responders using computer vision intelligence have the advantage of viewing video capture of the incident that is unfolding in real-time to contextually understand the scenario and prepare for proactive engagement—including during gun-related incidents.”

Seeing What’s in Your Pockets

Not everyone is in favor of AI scanners. Syracuse University professor Johannes Himmelreich who studies the ethics of artificial intelligence told Lifewire in an email interview that “the idea of fixing the gun problem in this country with AI is just terrible.” He said this proposal tries to apply a technological fix to a social problem. “Such attempts are generally misguided,” he added. “Even worse: they take the oxygen out of pertinent solutions.” Another problem that comes with using AI for any type of screen is equity, he said. “What we don’t want to see again is people of color being wrongly accused of carrying again at much higher rates than others. In principle, AI can reduce bias. But in practice it usually compounds such discrimination.” There’s also the matter of overreach. If AI-powered scanners gain in popularity they might be used to record people all the time, Pisani pointed out. “So not only would they record if they are carrying guns or not, but also their behavior,” he added." Every second of your life would be recorded, and there will be people that would be uncomfortable by this technology. When utilizing this technology, it is important to consider all the privacy implications and educate people on the impacts this technology can have." Evolv did not immediately respond to a request from Lifewire seeking comment.  But Nilay Parikh, the CEO of Be Global Safety, which makes AI-powered monitoring software, defended the use of AI. He said that in public places where guns may be prohibited and cameras already exist it is not a privacy invasion to use AI to detect guns.  “AI can have the ability to protect gun holders’ identity or can aid law enforcement in identifying the suspect,” he added.