Back in January, a group of Amazon shareholders put forward a proposition to ban the sale of the company’s facial recognition software, Rekognition, to government agencies around the world. For months the company has fought the proposal, using rules set by the Securities and Exchange Commission (SEC) to avoid the vote. But with the SEC ruling against the company and in favor of its shareholders just a few weeks ago (on March 28 and April 3), the tech giant was, for once at least, fighting a losing battle.
Now, the company is finally facing up to the reality of its position. Last week it confirmed the vote will take place at the annual shareholder meeting on May 22.
It might seem strange that shareholders would want to reduce a company’s ability to sell more products. But Rekognition is no normal product and has faced scrutiny for months from a number of different quarters.
Earlier this month, a group of prominent industry and academic AI researchers urged Amazon in an open letter to stop selling its facial recognition technology, Rekognition, to law enforcement agencies. The researchers argued that repeated studies and scrutiny have shown that Rekognition has higher error rates for people of color and women.
The sale of Rekognition to governments until Amazon could prove that “the technology does not cause or contribute to actual or potential violations of civil and human rights.”
There’s a good deal of data backing up the AI researchers’ position. In 2018 an ACLU study did an experiment in which Rekognition falsely matched 28 members of Congress to mugshot photos, while earlier this year an MIT Media Lab study showed that Rekognition struggled when trying to correctly identify women of color.
In response to this research, Amazon has gone on record asserting that researchers weren’t using Rekognition properly. However, that doesn’t really solve the issue: police say they don’t use Rekognition in the way Amazon claims it recommends, either.
It’s important to note that shareholders are not unwavering in their opposition to Rekognition. Instead they have called for “an evaluation using independent evidence” to ensure “the technology does not cause or contribute to actual or potential violations of civil and human rights.”
Essentially, the shareholders’ position is one about guaranteeing that the company is making the right level of checks and balances when developing a product that has the potential to be so transformative.
Amazon’s Board of Directors opposes the proposal. The Board cites the fact that Rekognition has never received any complaints about the product in relation to privacy or human rights from any of its customers.
In addition to the proposal to ban the sale of facial recognition tech to government agencies around the world, shareholders will also vote on another proposal to have a third party study the privacy and human rights implications of Amazon’s Rekognition.
The Amazon Board opposes that proposal too. It argues that their commitment to privacy and human rights is enough, outlined in the company’s Terms of Service. This has been the company’s position for the last year whenever it has been met with criticism.
Amazon’s perspective isn’t one that’s shared across the tech world. Last year, Microsoft’s president and Chief Legal Officer, Brad Smith, argued for for the federal regulation of face recognition:
“We live in a nation of laws, and the government needs to play an important role in regulating facial recognition technology,” Smith wrote. “As a general principle, it seems more sensible to ask an elected government to regulate companies than to ask unelected companies to regulate such a government.”
Amazon Rekognition faces more scrutiny from Democrats and German antitrust probe
AWS updates the face detection, analysis and recognition capabilities in Amazon Rekognition