A camera view of people standing in a city square being identified with facial recognition technology.
An arrow pointing leftHome

Amazon and Microsoft claim not to have used AI training data they’re being sued for exploiting

  • Mike Pearl

The legal filings in Seattle relate to a case over an Illinois law.

A pair of legal filings from May — one by lawyers for Amazon, and another by lawyers for Microsoft — add to the claims from the two tech giants that a privacy lawsuit filed back in 2020 is groundless. These latest court filings were unearthed earlier this month by Todd Bishop of Geekwire, who draws attention to the need for a federal privacy bill of rights, and the seemingly absurd legal machinations this gap makes possible — or perhaps necessary.

To be clear, what’s being alleged has serious implications: Microsoft and Amazon both, it seems, explored the use of a collection of facial images, and the plaintiff says they ran roughshod over an Illinois law aimed at protecting the exploitation of people’s biometric data in the process. The new filings claim the data they’re being sued for using was never actually put to use.

A legitimately explosive exposé by NBC from 2019 first drew attention to the story. A source leaked an IBM dataset to the network news outlet, and it revealed a large number of Flickr users whose photographs had been secretly included — and their inclusion had never been disclosed. IBM’s stated intention in creating the dataset had been the correction of racial and gender bias in data used to train facial recognition algorithms. The Flickr users in question had published their work under Creative Commons licenses, meaning it was free for certain uses. But, critically, inclusion in corporate facial recognition algorithms was apparently not one of them.

By the time the NBC report emerged, workers at Amazon’s computer vision platform “Rekognition” and Microsoft’s Photos app had both apparently downloaded the data (as did Google, incidentally). So two Illinoisans, Steven Vance and Tim Janecyk, who were unhappy to have had their faces used in this way, sued. They allege that the use of their facial data for this purpose violated an Illinois law called the Biometric Information Privacy Act (BIPA), which forbids images of faces or any other biometric information to be used without the consent of the subject for profit. Their original suits demanded 5,000 per violation.

“Facial recognition technology — once a thing only seen in movies — now threatens to end individual privacy. Public and private entities increasingly deploy facial recognition products to determine a private citizens’ identities, as well as other personal information, such as their addresses, phone numbers, whereabouts and acquaintances,” says the text of the original lawsuit against Microsoft.

A little over a year ago, Seattle federal court judge, James L. Robart, gave the Amazon and Microsoft suits the go ahead to proceed.

But if the May motions to dismiss from Amazon and Microsoft are to be believed, the vast and powerful tech companies are innocent victims. Amazon and Microsoft employed researchers who were curious to know if the IBM data — which was being provided for free for research purposes — would help eliminate racial and gender bias in their products. But, independently of one another, they both claim to have determined that it wouldn’t have, and opted against using it. So it was never actually used for profit.

And that’s all that ever happened with the data, they claim. The filings also spend a great deal of time pointing out the geographical oddities of the case: They’re Seattle companies, being sued for violating an Illinois law, by buying data from a New York company.

The lawsuits are seeking class-action status, and purport to represent the thousands of people victimized by this use of their facial data.