On January 27, 2022, we reported that the IRS had implemented facial recognition policies, compliance mandated by June 01, 2022, that would require online IRS services users to provide video selfies to the IRS to conduct activity on their site. Whether filing tax returns or requesting prior filing records, the usual drop-down identity verifying menu would no longer suffice to access IRS online services. Those wishing to avail themselves of these services would have to submit a video selfie to digital identification services provider, ID.me, for authentication.
On May 11, 2022, the IRS scrapped the selfie mandate due to opposition from privacy advocates and a bipartisan group of lawmakers that have since asked the Federal Trade Commission to investigate whether identity verification company ID.me illegally misled consumers and government agencies over its use of controversial facial recognition software. Basically, ID.me had reassured the IRS that they would be using 1:1 matching facial recognition technology (FRT) but was found to actually be using 1:many (1:N) FRT.
The two types of facial recognition matching are:
• One to one (1:1) matching -the face in the image is presented to the system and is compared with the face of a known person in the enrollment database to predict if they are the same.
• One to many (1:N) matching – the face in the image presented to the system is compared with the known faces in the enrollment database to see if any matches are found.
One major issue surrounding 1:N (one to many) FRT is that there are greater incidents of matching errors due to the flaws in the current imperfect technology. The face databases maintianed by most private FRT companies and the government contain many more Caucasian male face arrays than those of women and minorities. (Clearly, face accumulation databases are in direct proportion with current racial demographics in the United States.) Lawmakers pointed to the body of research demonstrating that facial recognition systems are often built with inherent racial bias that makes the technology far less accurate for non-white faces.
Also, concerns were raised that allowing a private company to collect face data from millions of Americans posed a cybersecurity risk.
As of today, both the IRS and ID.me have provided additional options that give taxpayers the choice of opting in to use ID.me’s service or authenticating their identity via a live, virtual video interview with an agent. Still, a very flawed policy by the IRS and with technology that has simply not undergone sufficient testing before going live. We can reasonably expect massive wait times, delays and confusion.
We will continue to monitor the various merging uses for FRT, especially when employed by government agencies, compounded by the reliance of these agencies on private sector virtual identification verification companies.
(Relatedly, now I’m curious about the use of facial recognition technology via the government’s employment verification program, E-Verify, a Department of Homeland Security site that allows businesses to determine employment elgibility in the United States both for U.S. citizens and foreign workers. I’ll return with my research results shortly.)
BNI Operatives; Situationally aware.
As always, stay safe.