Congressman Cohen Expresses Concerns about Issues of Inaccuracy with Facial Recognition Technology

WASHINGTON – Congressman Steve Cohen (TN-09), a member of the Science, Space and Technology Subcommittee on Research and Technology, today questioned Dr. Walter G. Copan, Undersecretary of Commerce for Standards and Technology and Director of the National Institute of Standards and Technology (NIST), about the origins of inaccuracies in some facial recognition algorithms. Last December, NIST found significant differences between the accuracy of algorithms and their ability to properly identify people of color.

Congressman Cohen asked:

"In light of all the news around Clearview AI and its secretive facial recognition system, I wanted to discuss NIST's important work in accuracy benchmarking for facial recognition technology through its Face Recognition Vendor Test program. As the debacle around Clearview AI shows us, this technology poses significant societal risk, and understanding their accuracy is paramount. In its most recent test last December, NIST found vast differences between the accuracy of algorithms, with the top 17 performing algorithms being nearly perfect across demographics while the bottom performing algorithms showed significant false-negative and false-positive rates.

"What accounts for the significant differences in accuracy rates between the top-performing facial recognition algorithms and the lowest-performing algorithms?"

See Congressman Cohen's exchange with Copan here.