Amazon’s Hopes for Rekognition: Amazon to pitch its facial recognition technology regulation to lawmakers
October 10, 2019
2 min read
What's going on here?
In September 2019, Amazon CEO Jeff Bezos unveiled the tech giant’s plans to develop its own draft legislation to regulate facial recognition technology. Amazon hopes to pitch this to lawmakers and for them to adopt its as many of its suggestions as possible.
What does this mean?
Since 2016, Amazon has been optimising “Rekognition”, a deep-learning artificial intelligence (AI) software which analyses images and videos to identify the objects, people, and text present. The software’s primary application is facial analysis and facial recognition. Using the software, Rekognition’s business partners will be able to match people’s faces with photos and videos from other databases in real time.
The move to pitch draft legislation to lawmakers follows on from Amazon’s blog post back in February 2019 which suggested ethical guidelines on the use of the controversial technology. Suggestions included a compulsory manual human review in certain circumstances (for example, if the tech was used in law enforcement) to ensure that its use would not result in the violation of a person’s civil rights due to an inaccurate prediction.
What's the big picture effect?
Amazon has kickstarted the move for greater regulation for AI of Rekognition’s kind, stealing a march on their rivals. In February, other tech giants such as Microsoft and Google also signalled the need for federal regulation. As tech moves much faster than the law on tech, it is essential that major tech companies call out for such legislative change (see our article on the call for more action on facial recognition here).
However, the vested interest Amazon has in the regulation it operates under will make some uncomfortable. There are fears that Amazon is prioritising profits above privacy as it seeks to shape the facial recognition market which is estimated to be worth $8 billion in 2022.
Controversially Amazon has marketed the software to law enforcement agencies as well as corporations. This is a development that has concerned the American Civil Liberties Union (ACLU) which is concerned about the possible risks of racial profiling and law enforcement abuse. Recent research by MIT which showed that commercial AI software tended to have higher error rates for black people evidences the risk of using AI in criminal justice. The research showed further that while light-skin men fared the best with an error rate of 0.8%, dark-skin women suffered with an error rate of 34.7%. Rekognition itself has been reported to have mismatched 28 members of Congress, disproportionately those of colour, with mugshots from criminal databases. With these concerns of potential abuse, a number of US cities (San Francisco and Oakland in California, Somerville in Massachusetts) have banned the use of facial recognition by government agencies for the foreseeable future.
It remains to be seen whether Amazon’s push for greater regulation will assuage critics’ fears or whether the move is more commercially motivated and so the changes made will not be as substantive as they need to be in order to safeguard consumers’ privacy. However, it is certain that, regardless of the motivations, Amazon’s move has once again brought facial recognition into the spotlight.
Report written by Heerim Hwang
Share this now!
Check out our recent reports!