When New York City Mayor Eric Adams announced the city would be testing out AI gun detectors on at subway stations last week, he touted the technology as “clearly impressive.” But critics, including the Legal Aid Society, were immediately skeptical of the pilot — and as it turns out, they were right to be.

When the city installed Evolv scanners at a hospital in the Bronx in 2022, the machines frequently reported false positives, according to a public records request obtained by Hell Gate. In the seven months that the Evolv scanners were active, 50,000 of the 194,000 scans resulted in a positive alarm — 85 percent of which were false positives. Most of the actual positives were triggered by police officers; just 295 were triggered by civilians carrying knives, guns, or weapons categorized in the report as “other.”

Evolv claims its AI-equipped scanners use “safe, ultra-low frequency, electromagnetic fields and advanced sensors to detect concealed weapons.” The company’s CEO has said the machines can detect any type of weapon, from guns to bombs to large tactical knives. But the 2022 pilot — as well as industry testing and more recent probes by the Federal Trade Commission and the Securities and Exchange Commission, as well as a class-action lawsuit filed by shareholders — suggests Evolv’s scanners are far less effective than advertised.