Facial recognition software used to verify unemployment recipients reportedly doesn’t work well

The facial recognition program used by some two dozen US states to to verify people seeking unemployment benefits is working inconsistently, leading to many people being denied benefits or having their applications put on hold, Motherboard reported.

The identity verification service ID.me is intended to help reduce unemployment fraud, and uses biometric data and official documents to verify people. But according to Motherboard, some who have applied for unemployment have reported that ID.me has failed to identify them correctly, and that they have had difficulty reaching someone at ID.me to remedy the problem.

In a lengthy email response to The Verge, ID.me CEO Blake Hall said the company uses “1:1 Face Matching to match the selfie image to the photo on the government ID. This is similar to how Apple uses FaceID to unlock phones and analogous to how a TSA agent would compare your face to your photo ID at an airport.”

He added that the algorithms the company uses for Face Match “operate at ~99.9% efficacy.” Hall also said the company was not aware of “eligible individuals” who couldn’t verity their identity with its software, and that wait times for a video chat with a company representative “has been consistently under 30 minutes all week.”

With millions of people unemployed due to the coronavirus pandemic, state offices have dealt with a surge of claims over the past year. Some states reported a sharp rise in fraudulent claims last spring, and the Department of Labor reported (.pdf) in February that between March and October of last year it had identified more than $5 billion of potentially fraudulent unemployment payments.

Update June 20th 9:24AM ET: Added comment from ID.me CEO

https://www.theverge.com/2021/6/19/22541427/facial-recognition-software-verify-unemployment-benefits-id-me