First U.S. transit system to use AI to search for weapons

On your next visit to Las Vegas, there will be an extra set of eyes watching you if you decide to use the local transit system.

As part of a $33 million, multi-year upgrade to enhance its security, the Southern Nevada Regional Transportation Commission is preparing to add a system-wide artificial intelligence system from gun-detection software company ZeroEyes that will scan passengers on more than 400 buses in an attempt to identify anyone carrying a firearm.

Tom Atteberry, director of operations for safety and security at the Firearms Control Board, said seconds matter in active shooter situations and implementing the system could give authorities an advantage. “Time is of the essence,” he said. “It gives us time to identify the firearm being brandished so we can notify them, get to the scene and save lives.”

Monitoring and preventing mass shootings is something that public places across the country struggle with every day. Violent crime on transit systems, in particular, continues to be a problem in major metropolitan areas, with the Department of Transportation releasing a report in late 2023 outlining the concerns of transit agency officials across the United States about rising violence on their transit systems. According to a database maintained by the Bureau of Transportation Statistics, attacks on transit systems have been on the rise, and there has been a rise in public concerns about transit safety.

Purdue University Northwest, which has about 7,000 students, has students of all ages arriving from urban areas such as Gary and Hammond, Indiana, by bus, car and train. On average, from the first gunshot to the first 911 call, it takes five minutes. On average, it takes 12 minutes before police can “put their boots on the ground” at a school, said Brian Miller, the university’s director of public safety, citing research on response times to mass shootings. “We need to reduce that, and gunshot detectors and gun scanners are all good technologies,” Miller said.

While officials in Nevada and Purdue University Northwest said there has been no specific incident that sparked interest in the new AI-powered detection technology, the risk of an accident is always on the minds of public safety officials. Las Vegas transit crime statistics are low compared to similarly sized transit systems, but the issue of guns in public places has been at the forefront of the conversation since Stephen Paddock killed 58 people on The Strip in a 2017 mass shooting.

Mass shootings typically begin with someone entering an apartment complex with a weapon on full display, giving ZeroEyes room to operate and navigate a fog-of-war-like situation, according to Sam Alaimo, co-founder of Zero Eyes. Typically, as mass shootings unfold, there are multiple 911 calls from witnesses, all providing conflicting information.

If ZeroEyes detects a firearm, it alerts an operations center staffed by retired law enforcement officers, who quickly determine whether the weapon poses a threat.

Atteberry stressed that the ZeroEyes system is not designed to detect legal firearms that may be holstered or packed in a purse. “It’s only designed to identify someone brandishing a firearm in a threatening manner,” he said.

The process from the moment the image arrives, to their staff checking it, to alerting 911 can take no more than three to five seconds, Alimo said. The system can usually identify the type of weapon as well, which he added could give law enforcement an advantage.

“We give them the clarity to go in there and stop the killing,” Alaimo said. “Their goal is to cause mass damage. When you see these mass shooters, they usually start outside, waving their guns around, showing their weapons. We built this building with the mass shooter in mind.”

ZeroEyes took off as the pandemic hit. Alaimo said the company was designed specifically for school campuses, but with schools closed, the company has shifted to government buildings, corporate campuses, casinos and industrial parks.

Acoustic sensors compete for the same public safety market.

Image-based scanning software like ZeroEyes, which relies on existing camera mounts, may not be as effective as other AI options at detecting weapons, says Wei Dai, an assistant professor in the Department of Computer Science at Purdue University Northwest and director of the Advanced Intelligence Programs Laboratory. Audio sensors may be more appropriate in some cases because cameras can’t cover every inch of a building or campus, Dai says, but audio sensors detect the sound of a gunshot with 99 percent accuracy.

“AI comes from data; if we can’t capture the data, there will be no successful AI,” Dai added.

But acoustic sensors have been slow to catch on, largely because of cost. Earlier this year, Seattle scrapped a plan to install acoustic sensors in high-crime areas. But the devices are being used in some areas. Lafayette, Louisiana, is testing them this year, and Clark County, Nevada, home to Las Vegas, is launching a pilot program to use acoustic sensors, though not in its transit system.

Purdue University Northwest’s public safety team, which has 14 full-time officers and uses a variety of tools to keep itself safe, will look at all technology options in the future, including image-based technology like ZeroEyes. “As technology advances, that’s something we might be interested in,” Miller said. “You need a multi-layered approach to law enforcement.” While photos, audio and other technologies, combined with old-fashioned policing, provide the best comprehensive approach, he added, there is no single tool to prevent mass shootings.

Las Vegas’ transit system isn’t the only one in the country using ZeroEyes, but it’s the first to plan for full deployment. A pilot program with SEPTA in Philadelphia ended this spring after a year, but Paul Gratton Jr., a former NYPD superintendent with the Bureau of Transportation who now works as a consultant, says Las Vegas could be a better test case.

“ZeroEyes offers the AI ​​part, but they don’t install their own cameras. SEPTA found that most of their existing cameras weren’t of sufficient quality to enable the AI ​​to do the job. I think Vegas is a much better platform; their camera system is better,” Gratton said.

Alaimo says ZeroEyes has value wherever a newer camera network is already in use.

Even with better cameras, implementing the technology involves many moving parts, such as what to do once a firearm is detected. “What actions are you allowed to take, and what actions are responding officers required to take?” Gratton said.

While image-based AI detection systems are not perfect, they are useful in a comprehensive security package and are already widely used in intelligence surveillance. Gratton believes technologies like ZeroEyes will be adopted much more widely, and that cameras are well-suited to transportation systems because they can handle large-scale screening. “You can’t subject everyone to the level of security screening that the TSA requires,” Gratton said. “Passengers will only tolerate a limited amount of screening.”

“We’ve tested a lot of similar technologies, and I think the future will be technology like ZeroEyes,” he said. “These camera systems are the natural evolution of policing that will include artificial intelligence in the camera systems.”

Cybersecurity consultant Michael Haas remains skeptical, citing the history of gun hiding, which is in effect a cottage industry.

“Concealed weapons have been an art form for hundreds of years,” Haase said. “Such systems work best when there is no attempt to conceal the weapon, but the simple fact is that there are many ways to conceal a weapon, and all of these systems rely on technology that can be easily fooled, from completely plastic weapons to simple modifications to the form factor.”

But Attebery says the stakes are too high to ignore the new technology. “You have to be progressive and proactive; you can’t just sit back and wait for something to happen and then act,” he says. “Our job is to look ahead and prevent something from happening. If it saves one life, it pays for itself.”