Lucy McBath Asks AI Experts About Threats To Fourth Amendment Posed By Facial Recognition Technology

Forbes Breaking News

During a House Judiciary Committee hearing on Wednesday, Rep. Lucy McBath (D-GA) spoke about the Fourth Amendment.

Transcript

00:00And now I recognize Ranking Member Ms. McBath.

00:04Thank you, Mr. Chair.

00:05And I just have to say that, you know,

00:07and just in listening to each and every one of the witnesses,

00:10I mean, I'm just really amazed at the depth of the use,

00:16criminal usage of AI.

00:18Really, thank you so much for what you bring to the table today.

00:22But I do kind of want to talk a little bit about facial recognition.

00:25The Detroit Police Department reportedly conducted 129 facial recognition

00:31searches in 2020, and all on African-American people.

00:36And the following year, 95.6% of the searches targeted black people.

00:44Mr. Vinsky, how does the use of AI-enabled facial recognition comply

00:49with the Fourth Amendment's equal protection principles?

00:53It raises serious concerns.

00:55And as far as I know, there's not been a clear holding

00:58that equal protection principles are violated

01:00by the use of facial recognition technology.

01:02But it certainly raises those concerns

01:04because of the disproportionate impact we've seen that technology

01:08have on protected classes, particularly, as you said,

01:12black people and, in particular, black men.

01:14And we've also engaged in civil rights litigation

01:17to defend individuals who have been wrongly identified

01:21by facial recognition technology.

01:23To a large extent, this is a matter of process,

01:26ensuring that police departments have appropriate processes in place

01:29so that there isn't reliance solely on an identification

01:34made by facial recognition technology

01:37to bring in a suspect that there isn't bias in lineups

01:41and things of that nature.

01:42But because of the certain threats that we've seen here

01:46and the ways that the technology can struggle

01:49in real-world conditions,

01:51we have long stood by that there needs to be a moratorium

01:53for law enforcement uses of federal recognition technology.

01:57Thank you for that.

01:59And it's just so interesting you should say that

02:00because even on my phone, I have facial recognition,

02:04and sometimes it says, I don't recognize you.

02:07And I'm like, well, you know who I am.

02:09But, of course, there's still problems with AI,

02:12and the technology still needs to be advanced.

02:14Mr. Vinsky, I'm going to ask you another question.

02:16What warrant requirements and limitations

02:21should be applied for facial recognition tools

02:23when they're used by law enforcement?

02:25Well, as I said, our overall stance is that law enforcement

02:29should not be deploying the technology at all

02:31because of the underlying foundational issues

02:34of how it can struggle with a variety of protected classes

02:38and correctly identifying people,

02:40especially in real-world conditions

02:42where lighting may not be ideal,

02:45where the surveillance footage may be grainy,

02:47and that can result in someone who is eight months pregnant

02:50being apprehended for a crime

02:51where there clearly was not a pregnant person involved.

02:54The use of facial recognition technology

02:59may not necessarily implicate the Fourth Amendment,

03:01but as I said, it raises very serious concerns

03:04about perpetual surveillance,

03:06the ability of the federal government

03:07to identify individuals in public spaces

03:10going about their daily lives

03:12without any recourse, without any judicial oversight,

03:16and that is ultimately a policy question

03:18for legislators, city councils, and Congress

03:21to step up and regulate.

03:22Thank you for that.

03:24And Dr. Bowne, in your experience,

03:26have you found that AI-enabled tools used by law enforcement

03:30are tested and evaluated before they are deployed

03:33to ensure that they are safe and effective?

03:36So in my experience in law enforcement as a prosecutor,

03:46recently as a supervising prosecutor,

03:49the tools that are being used

03:52are certainly going to depend on the jurisdiction

03:56that's using them.

03:57And so states, counties, certainly federal government,

04:05from my experience in the Department of the Air Force,

04:09law enforcement organizations

04:11that are starting to use these tools

04:13are relatively new users.

04:17In the Air Force, any AI tool is supposed

04:20to go through rigorous test and evaluation standards

04:23to ensure that it results in a certain quantified reliability.

04:31And that's very challenging to do with some of these tools,

04:34particularly when you're talking about edge cases

04:37like African-American men

04:38when they are not found frequently in data sets.

04:42And so those questions are being asked.

04:45I don't see, in my experience,

04:47the type of rigorous standards

04:51being established across the board by regulators.

04:57And so until that happens,

04:59law enforcement is going to try to keep up,

05:02and it's tempting to do that,

05:03but there's likely to be some gaps there.

05:09General Lady Yield.

Category

Transcript

Recommended