Because Facial Recognition Tech Just Isn’t Sketchy Enough, Cops Are Now Running Searches Using AI-Generated Faces

(Mis)Uses of Technology

from the two-wrongs-in-search-of-a-right-(to-violate) dept

Facial recognition tech is probably improving as time goes on. Given enough providers, controversy, and individuals who definitely want this tech to stop being so terrible at correctly identifying women and minorities, anything is possible.

Rather than wait for the tech to catch up to the promises made by promotional materials, cops are apparently moving ahead with efforts that will cause even more problems for future tech adoption.

We’re already afflicted by at least one tech company that believes it’s perfectly OK to stock its database of billions of photos with any content not locked down on the internet. Beyond that, there’s the problems inherent to the systems themselves, which aggravate biased policing by doing their most accurate work when gazing on the faces of white males.

Now there’s this, which is the sort of thing that’s just a lawsuit waiting to happen. Here’s Paige Collins and Matthew Guariglia of the EFF with more details:

A police force in California recently employed the new practice of taking a DNA sample from a crime scene, running this through a service provided by US company Parabon NanoLabs that guesses what the perpetrators face looked like, and plugging this rendered image into face recognition software to build a suspect list.

Parabon has been offering its DNA-to-face services for years. It’s very proud of its ability to generate faces using nothing but DNA info. It has tons of cases listed on its site and provides links to news coverage of investigations aided by its ability to generate a DNA-based analogue for police suspect sketches.

Perusing the site, it’s immediately noticeable that lots of the DNA-based speculations look very little like the person arrested or charged. But that’s not really all that problematic. Parabon’s “snapshots” aren’t meant to be definitive descriptions of criminal suspects. They’re simply meant to contribute to ongoing investigations by giving cops something to post or hand out when asking people if they’ve seen anyone resembling these speculative pictures.

Sure, there’s always a chance this may result in a wrongful arrest or detention, but the company makes it clear these are nothing more than a best guess based on DNA profiles. It’s not great, but it’s not Parabon’s fault if cops decide to go a step or two further than the purpose for which these “snapshots” were intended.

That’s on the cops themselves. Parbon does not encourage this sort of use of its DNA snapshots. But cops who apparently have zero concern about adding AI speculation to AI speculation to engage in investigations are making things demonstrably worse by using Paragon’s “snapshots” for reasons they were never intended. Wrong + wrong never equals right.

This puts a second layer of speculation between the actual face of the suspect and the product the police are using to guide investigations and make arrests. Not only is the artificial face a guess, now face recognition (a technology known to misidentify people)  will create a “most likely match” for that face.

We already know facial recognition tech is flawed. This more than doubles the magnitude of unintended consequences by feeding pure speculation to an algorithm that likely already has problems accurately identifying people, especially if those people are minorities or women. That it might perform better on white males doesn’t matter much when it’s just questionable AI being examined by other questionable AI.

As the EFF notes, the cited search was a violation of Parabon’s terms of service. But cops rarely care if they violate laws, much less a private party’s rules of engagement. Any case closed by violating rights and/or terms of service will be treated as a success story. Every failure will simply be considered the acceptable cost of doing police business, even if the failure results in an expensive lawsuit settlement.

What’s exposed here is the tip of the iceberg. If one cop shop is doing this, the odds are several have done the same thing. The only difference is they haven’t been caught yet. Coverage like this may make it clear cops prefer power to responsibility, but the best deterrent remains meaningful consequences for their careless actions. And, to date, this country has shown — via law enforcement agencies and the judicial system that handles lawsuits resulting from their abuses of power — it will almost always consider it better to suffer abuse than actually hold cops accountable.

Filed Under: ai, dna to face, evidence, facial recognition, police
Companies: parabon nanolabs