At least eight suspects have been arrested on the basis of AI recognition software and later let go, according to a report.

The two latest cases involve Christopher Gatlin, a Missouri father of four, and Jason Vernau of Miami, according to the Washington Post. Gatlin was arrested after a security guard was assaulted by two men on a train platform in St. Louis.

Police only had a blurry photo from camera footage to go by, and the assailant in the photo was wearing a hood and a surgical mask. Despite this, police fed the photo into AI recognition software, which matched the photo to several mug shots, including one of Gatlin.

The man who was assaulted could not identify his attackers due to brain damage suffered in the assault, the outlet reported. The police officers investigating the case brought two mug shots to the injured security guard, Michael Feldman, who maintained that he could not remember any details but eventually identified the mug shot of Gatlin as being the matching photo.

Two years later, Gatlin was cleared of charges, according to the report.

The outlet found that while the police department’s AI policy warned that AI software is “nonscientific” and “should not be used as the sole basis for any decision,” it and 15 departments across 12 states had arrested suspects based on AI recognition software.

CLICK HERE TO READ MORE FROM THE WASHINGTON EXAMINER

Of the eight suspects identified in the report, the police failed to check alibis in six; ignored contradictory evidence in two; failed to collect other evidence in five, such as the case of Vernau, who was accused of cashing a fraudulent $36,000 check at a Miami bank; and relied on problematic witness statements in six.

In one case, Porcha Woodruff was accused of committing a carjacking despite being seven months pregnant at the time of the crime. No witness identified a heavily pregnant woman being at the scene, according to the report.