How AI could be used to detect guns before school shootings

• Yes, but: Critics say AI surveillance systems aren't effective, and they worry about other issues, including the technology's record of bias, particularly against people of color.

Driving the news: Many Republicans, including former U.S. President Trump and Texas Sen. Ted Cruz, have called for tougher school security measures, including single points of entry, metal detectors, more school police and armed teachers.

• But metal detectors can clog entrances, which could make students more vulnerable to attack, security experts note.
• Security cameras aren't routinely monitored — they're more likely to be a source of evidence after an incident. And gunshot-detection technologies like ShotSpotter, which uses acoustic sensors to identify loud noises, have been criticized as ineffective.

How it works: Some companies are promoting AI technologies as a less obtrusive and more effective alternative to metal detectors and other systems.

Evolv uses a combination of ultra-low frequency electromagnetic fields and advanced sensors to detect concealed weapons as people walk through a portal.

• The high-speed screening system is already in use at several sports arenas, entertainment venues and theme parks, the Washington Post reported.
• North Carolina’s Charlotte-Mecklenburg school system, with 150,000 students, is installing the scanners at a cost of $1.7 million over three years.
• It's not foolproof, however: Evolv's system generated false positives from certain Google Chromebook laptops, according to an analysis by IPVM, a security-industry trade publication.

Hexwave, developed at a Massachusetts Institute of Technology lab and licensed by Liberty Defense Holdings, is a similar system. It uses 3D imaging and AI to detect concealed weapons as people walk between two panels.

• It can detect both metallic and non-metallic objects like 3D printed guns, and plastic or liquid explosives, but is also trained to recognize ordinary objects like cell phones, wallets or keys, CEO Bill Frain tells Axios.
• Trials will begin in August at five locations: Toronto's Pearson International airport, the Port of Tampa's cruise ship terminal, Baltimore's Oriole Park at Camden Yards, a major Hindu temple near Atlanta and the University of Wisconsin.

ZeroEyes, used by schools in 18 states and now being piloted at Oxford High, integrates its AI software with a school's existing surveillance cameras to identify guns in a camera's field of view. (Assailants often brandish their weapon before a rampage, the company says.)

• The security image, stamped with the time and location, is then reviewed by a trained military veteran to verify that the object detected is in fact a gun.
• Within about five seconds, ZeroEyes will issue an alert to school authorities, providing a description and location of the threat — often before any shots are fired, cofounder Mike Lahiff tells Axios.
• The technology serves as "a force multiplier" with the goal of getting better information to first responders more quickly, he said.

The big question is whether these systems will prevent future shootings.

• ZeroEyes says it has detected "thousands of guns" since launching in 2018, including BB guns and paint guns that students use to play games like "Assassin" on school grounds.
• But AI technology is expensive — $25 monthly per camera, in the case of ZeroEyes. In a large high school with 200 cameras, that's $5,000 per month.
• Schools don't have that kind of money, says Jason Russell, a former U.S. Secret Service agent and founder of School Education Consultants, whose company vets technologies for school districts.
• While he likes ZeroEyes' human-in-the-loop technology, if it works well, it's subject to future budget cuts. "That's the paradox of security: When you do it really well, and nothing happens, it appears unnecessary."