• The error rate for light-skinned men is 0.8%, compared to 34.7% for darker-skinned women, according to a 2018 study titled “Gender Shades” by Joy Buolamwini and Timnit Gebru, published by MIT Media Lab. A 2019 test by the federal government concluded the technology works best on middle-age white men. The accuracy rates weren’t impressive for people of color, women, children, and elderly individuals.
• We sued on behalf of Kylese Perryman, an innocent young man who was falsely arrested and detained based solely on incorrect facial identification.

• You unlock your phone using Face ID. A traffic cop identifies a suspect with real-time surveillance footage. Retailers use it to track in-store customer behavior. These aren’t scenes from a sci-fi movie. It’s just another Tuesday in the age of facial recognition, powered by image recognition technology.
• racial bias in algorithmic accuracy, surveillance overreach, lack of transparency, and troubling questions about consent.
• Racial bias and discrimination  Misidentification of people of color, wrongful arrests, systemic inequality  Train models on diverse datasets; conduct independent audits; mandate bias testing  Data privacy  Unauthorized data collection, surveillance, and misuse of sensitive biometric data  Enforce opt-in consent, minimize data collection, and strengthen storage protections  Lack of informed consent and transparency  Use of facial data without user awareness or permission  Standardize consent processes, regulate dataset sourcing, and ensure disclosure policies  Mass surveillance  Loss of anonymity, chilling effects on expression, unchecked state monitoring  Restrict public deployment, require oversight, and enact legal safeguards  Data breaches Identity theft, data leaks, and limited recourse for affected individuals  Encrypt facial data, enforce breach disclosure, and establish stronger biometric laws
• • Who is harmed? People of color, the elderly, and individuals already overrepresented in policing databases.
   
  • What are the consequences? False arrests, increased surveillance, biased policing, and systemic discrimination.
• half of American adults, or nearly 117 million people, have photos on law enforcement's facial recognition network. However, it’s disturbing that errors detected in the face recognition system were more common on dark-skinned faces, but fewer errors were made when matching light-skinned faces.
• It reported that facial recognition technologies for 189 algorithms showed racial bias toward women of color.

• 35%

   

  of facial recognition errors happen when identifying women of color, compared to 1% for white males.

   

  Source: Massachusetts Institute of Technology

5 more annotations...

• work chronicles the story of Clearview AI, a small, secretive startup that launched an app in 2017, using a 30-billion-photo database scraped from social media platforms without users’ consent.
• Clearview’s software is used in crime investigations and has been used by law enforcement to identify Jan. 6 rioters, said Hill. Facial-recognition technology can also be used to intimidate or harass investigative journalists, government officials, or political opponents, she said. It can also lead to wrongful arrests, and she said measures should be put in place to regulate its use.

• stronger safeguards on the use of facial recognition technology after the Home Office admitted it is more likely to incorrectly identify black and Asian people than their white counterparts on some settings.
• latest testing conducted by the National Physical Laboratory (NPL) of the technology’s application within the police national database, the Home Office said it was “more likely to incorrectly include some demographic groups in its search results”.
• NPL’s finding “sheds light on a concerning inbuilt bias” and urged caution over plans for a national expansion.
• It can be used while examining live footage of people passing cameras, comparing their faces with those on wanted lists, or be used by officers to target individuals as they walk by mounted cameras.

• “the false positive identification rate (FPIR) for white subjects (0.04%) is lower than that for Asian subjects (4.0%) and black subjects (5.5%)”.
• false positives for black women was particularly high. “The FPIR for black male subjects (0.4%) is lower than that for black female subjects (9.9%),” the report said.
• “This has meant that in some circumstances it is more likely to incorrectly match black and Asian people than their white counterparts. The language is technical but behind the detail it seems clear that technology has been deployed into operational policing without adequate safeguards in place.”
• “The racial bias in these stats shows the damaging real-life impacts of letting police use facial recognition without proper safeguards in place. With thousands of searches a month using this discriminatory algorithm, there are now serious questions to be answered over just how many people of colour were falsely identified, and what consequences this had.

6 more annotations...