Letter to Nancy Pelosi, Speaker of the House and Kevin McCarthy, House Minority Leader - Rush, Tlaib Call on House Leadership to Prohibit the Use of Federal Funds to Purchase or Use Facial Recognition Software
Dear Speaker Pelosi and Leader McCarthy:
As the House prepares to take up the FY2021 appropriations bills, we write you on a matter of
national significance. As you know, our country is finally addressing long-overdue injustices that
have plagued communities of color throughout the United States. One of the factors that has begun
to receive attention is law enforcement's use of facial recognition software.
Though this technology is becoming more prevalent in society, it remains severely error-prone and
particularly so when used to analyze people of color. Notably in July 2018, the American Civil
Liberties Union (ACLU) tested Amazon's facial recognition software by comparing Members of
Congress against publicly available arrest photos and found 28 Members were erroneously
identified as those in the arrest database. While the ACLU tested all sitting Members of Congress,
the results of their investigation found that "the false matches were disproportionately of people
of color, including six members of the Congressional Black Caucus."
While the ACLU study was an experiment, we have seen the real-world implications of the use of
facial recognition software. On June 24, 2020, The New York Times reported on the case of Robert
Julian-Borchak Williams of Michigan, who was arrested by Detroit Police because he was
misidentified by facial recognition software. Furthermore, as reported by Vice, Detroit Police
Chief James Craig estimates that the software misidentifies individuals 96 percent of the time.
This, coupled with the fact that thus far in 2020, the Detroit Police Department, by its own
admission, has used facial recognition technology "almost exclusively against Black people" is
alarming and raises concerns that the technology is simply supercharging existing biases in
policing, disproportionately effecting communities of color and other overpoliced communities.
This situation isn't limited to Detroit or to the specific vendor used there. As a study by the
Massachusetts Institute of Technology (MIT) found, "when the person in the photo is a white man,
the software is right 99 percent of the time." That, however, is not the case for other racial groups
or for women. In fact, the MIT study found that the error rate is as high as nearly 35 percent for
darker skinned women. This study is not unique in its outcome. In its own study, the National
Institute of Standards and Technology found that facial recognition software "falsely identified
African-American and Asian faces 10 times to 100 times more than Caucasian faces."
Even if facial recognition technology were accurate, there are still broad concerns with its use.
Police surveillance cameras are disproportionately installed in communities of color, and these
same communities already suffer disparities at every stage of the criminal justice system.
Communities of color are more likely to be arrested, have improper force used against them,
receive longer prison sentences, and be subject to harsh confinement conditions. Facial
recognition risks making these problems worse, by arming law enforcement with a flawed tool
that is most likely to be targeted at communities of color. The aforementioned statistics released
by the Detroit Police Department simply confirm this fact.
Additionally, facial recognition provides unprecedented power for the government to track and
monitor people, whether attending a protest, place of worship, hospital, or social event, raising a
host of constitutional concerns. For example, in Baltimore, Maryland, police reportedly used facial
recognition on photos posted to social media to identify and arrest individuals who attended a
Freddie Grey protest, raising serious First Amendment concerns. We have already seen how this
technology can be further abused in examples from abroad, where foreign governments have used
facial recognition to repress critics and religious minorities.
The fact that such a powerful tool has been deployed largely in secret, without explicit
Congressional authorization, is alarming. What we do know, though, confirms that its current use
is not consistent with the United States Constitution. The Department of Justice reportedly has
access to over 640 million photos for face recognition matching and has used the technology
hundreds of thousands of times. Despite this, based on testimony in front of the House Oversight
and Reform Committee, the Department does not appear to be complying with its constitutional
notice obligations; does not track basic statistics, including the number of false identifications; and
does not even require probable cause or a warrant before running a facial recognition search.
It is clear that facial recognition technology is not reliable, and even if it were, would still raise a
range of serious concerns. Recognizing the fundamental dangers this technology poses, several
major companies, including Google, IBM, Amazon, and Microsoft, have already indicated they
will halt the sale of this technology to police. What is not clear, however, is how this corporate
decision will apply to federal agencies. Moreover, while these companies are well known to the
general public, they are not the major players in law enforcement facial recognition.
We firmly believe that the United States should not be using technology that is likely to mislabel
a significant portion of its population as criminals simply because of the color of their skin, and
which poses broader threats to civil liberties. For these reasons, we ask that the FY2021
appropriations bills include a prohibition on expending any federal funds, including grants to states
and localities, to purchase or use facial recognition software or contract for any similar services.
Thank you for your consideration of this request. If you or your staff have any questions, please
do not hesitate to contact us.