Trotz falscher Verhaftungen gibt die Polizei den Einsatz von Gesichtserkennung selten bekannt | Eine Untersuchung der Post ergab, dass sich viele Angeklagte der Rolle der Technologie bei der Verbindung zu Straftaten nicht bewusst waren, was zu Fragen der Fairness führte

Some key points from this investigative piece:

>Police departments in 15 states provided The Post with rarely seen records documenting their use of facial recognition in more than 1,000 criminal investigations over the past four years. According to the arrest reports in those cases and interviews with people who were arrested, authorities routinely failed to inform defendants about their use of the software — denying them the opportunity to contest the results of an emerging technology that is prone to error, especially when identifying people of color.
>
>In fact, the records show that officers often obscured their reliance on the software in public-facing reports, saying that they identified suspects “through investigative means” or that a human source such as a witness or police officer made the initial identification.
>
>…
>
>Defense lawyers and civil rights groups argue that people have a right to know about any software that identifies them as part of a criminal investigation, especially a technology that has led to false arrests. The reliability of the tool has been successfully challenged in a handful of recent court cases around the country, leading some defense lawyers to posit that police and prosecutors are intentionally trying to shield the technology from court scrutiny.
>
>Police probably “want to avoid the litigation surrounding reliability of the technology,” said Cassie Granos, an assistant public defender in Minnesota. This year, one of her colleagues helped persuade a judge to exclude a facial recognition result from the state’s case against an alleged thief because, the judge ruled, the software does not “consistently produce accurate results.”
>
>Misidentification by this type of software played a role in the wrongful arrests of at least seven innocent Americans, six of whom were Black, according to police and court records reviewed by The Post and reports in other news outlets. Charges were later dismissed against all of them. Some were told during interrogations or in documents provided to their criminal defense lawyers that they had been identified by AI. Others learned about the software’s use only after officers mentioned in passing that “the computer” had found them, or that they had been a “positive match.”
>
>…
>
>Federal testing of top facial recognition software has found the programs are more likely to misidentify people of color, women and the elderly because their faces tend to appear less frequently in data used to train the algorithms, according to Patrick Grother, who oversees biometric testing at the Washington-based National Institute of Standards and Technology. Roughly 2 million people of color and 2 million women are arrested in the United States each year, according to federal data.
>
>Clearview’s contracts with several police departments, obtained by The Post, say the program is not designed “as a single-source system for establishing the identity of an individual” and that “search results produced by the Clearview app are not intended nor permitted to be used as admissible evidence in a court of law or any court filing.”
>
>Prosecutors are required to inform defendants about any information that would help prove their innocence, reduce their sentence or hurt the credibility of a witness testifying against them. When prosecutors fail to disclose such information — known as a “Brady violation” after the 1963 Supreme Court ruling that mandates it — the court can declare a mistrial, overturn a conviction or even sanction the prosecutor.
>
>No federal laws regulate facial recognition and courts do not agree whether AI identifications are subject to Brady rules. Some states and cities have begun mandating greater transparency around the technology, but even in these locations, the technology is either not being used that often or it’s not being disclosed, according to interviews and public records requests.

This lack of disclosure for the use of these technologies is problematic at best, and especially so given the inaccuracies inherent in their algorithms and training data. Ideally systems that have these inaccuracies would not be deployed until these are remedied, but it seems that more often than not systems and services are rolled out with the expectation that the public will provide testing services, regardless of the broader consequences.