Ever AI under fire for using photo app images to train facial biometrics without adequate transparency

Photo storage app Ever used images uploaded by customers to create albums in the cloud were used to develop Ever AI biometric facial recognition technology, sparking allegations that the company violated user privacy, NBC News reports.

A brief reference to facial recognition training was added to the company’s privacy policy after NBC News contacted the company in April, according to the report, prior to which there was no notification provided to users that their images would be used to train biometric algorithms. Previously, the app’s policy had noted that facial recognition technology was used to help organize and share images, with an opt-in tagging feature.

Jacob Snow, American Civil Liberties Union of Northern California technology and civil liberties attorney, calls the move “an egregious violation of people’s privacy. Ever CEO Doug Aley told NBC News that Ever AI does not share user photos or personally identifiable information with facial recognition clients. He also noted the updated privacy policy, though NBC News says most of the seven Ever users it contacted were unaware that their images could be used to train facial recognition.

“I think our privacy policy and terms of service are very clear and well articulated,” Aley told NBC News. “They don’t use any legalese.”

New York University Law Professor Jason Schultz, however, said that including the notification in a 2,500-word long privacy policy is insufficient.

Ever AI says that it draws on a private dataset of 13 billion photos and videos to train its technology. The company also touted the accurate performance of its algorithms in NIST FRVT testing as making it a leading Western challenger to AI industry leaders in China and Russia.

NBC News notes that Ever AI markets its facial recognition to law enforcement agencies, which has also generated significant controversy for several companies, though its contracts so far are all with private companies.

Aley told NBC News that he joined the company in 2016, and that it shifted its focus to facial recognition shortly before its successful $16 million 2017 funding round.

How the data sets used to train and test AI systems are gathered has come under increasing scrutiny recently, with previous reports suggesting images used to train facial recognition are often collected without permission, and public revelations about the annotation of Alexa recordings.

Article Topics

artificial intelligence  |  biometrics  |  dataset  |  Ever AI  |  facial recognition  |  training