As part of the American Civil Liberties Union test, which identified 28 Congressional members as arrested by their recognition face recognition agency, some of these legislators have now raised serious concerns about the tech giant.
Amazon has not answered her questions or Ars.
"We ask for an immediate meeting with you to discuss how to fix the flaws in this technology to avoid inaccurate results," wrote Member of Parliament Jimmy Gomez (D-California). and Rep. John Lewis (D-Georgia), in a letter to CEO Jeff Bezos.
These two members of the House of Representatives were among the 28 who were falsely identified as being among a group of 25,000 mugshots.
Among the 28 were also Senator Edward Markey (D-Massachusetts), MP Luis V. Gutiérrez (D-Illinois) and MP Mark DeSaulnier (D-California), who separately phoned the company on Thursday. Answer specific questions and inquiries, with a deadline of August 20.
Please describe in detail how Amazon tests for face recognition accuracy, how often Amazon tests, and whether these results were independently verified.
Please describe in detail that Amazon conducts tests on bias in facial treatment. Appraisal results, in particular racial prejudice
Please provide a list of all law enforcement or intelligence agencies that (1) have contacted Amazon regarding the acquisition of Rekognition or otherwise has communicated with it and (2) is currently using the Recognition Service. In addition to your response to this letter, we encourage Amazon to include a list of Government Recognition customers in its next transparency report.
Amazon did not respond to Ars' question as to whether Bezos or any other business leader would meet these concerned members of Congress
Another legislator, MP Emanuel Cleaver (D-Missouri), who is not among the 28 Other issues were also raised:
Exactly @keithellison and I warned against this. Also, why I continue to be extremely worried that @amazon 's recognition will further aggravate the problems of racial prejudice in society, and especially when used by law enforcement agencies. https://t.co/8CgUxeIujn
– Rep. Emanuel Cleaver (@repcleaver) July 26, 2018
In a statement that Ars released Wednesday and other media on Thursday, Amazon said that his confidence threshold of 80 percent is not precise enough for human faces, a point the ACLU pushed back.
Jacob Snow, an attorney for the ACLU of Northern California, said in his own statement on Thursday:
In addition to these very real concerns, Congress members, community groups, and Amazon's own employees, shareholders, and consumers are repeating Amazon confirms the recognition – a product that aggressively markets it for law enforcement – can and will incorrectly identify people by default. This is downright dangerous, and there is more:
We know from our test that Amazon does not endeavor to ask users what they are using for recognition. Instead, Rekognition sets a standard: the same 80 percent we used to run our test.
We also know that the Amazon website is currently showing the use of 80% confidence in recognizing human faces. It shows that Amazon recommends an 80 percent trust in "Face-Based User Verification". If an 80 percent threshold is not appropriate for identifying people with a reasonable level of security, Amazon highlights this level of confidence Faces
On Thursday afternoon, Jesse Freund, an Amazon spokesman, contacted Ars to say that our earlier coverage of the ACLU report was "inaccurate". He refused, however, to explain further, without going "in the background", which we rejected. He also did not explain why he would not speak on the record.