The U.S. Rubber-Stamps Facial Recognition Systems Used to Round Up Uighurs in China
National Institute of Standards and Technology’s facial recognition audits are used for marketing purposes in China
Pegged as one of tech’s ultimate “disruptors” in 2019 by CNBC, Yitu Technology is a surveillance giant in China. The company has raised $400 million in venture funding, including from American VC firm Sequoia Capital, and has installed its facial recognition technology in 1,500 Chinese banks.
Yitu is also involved in more controversial facial recognition implementations. The company built a feature into its facial recognition software that was specifically meant to detect Uighurs, China’s persecuted Muslim ethnic minority. As a result, the company has landed on the U.S. government’s “entity list,” which means Yitu cannot buy products from U.S. companies without a special permit.
At the same time, Yitu’s marketing materials boast of a glowing evaluation from the U.S. government’s National Institute of Standards and Technology (NIST), which independently tests facial recognition algorithms for accuracy. Yitu Technology has pointed to its NIST scores for years. “YITU wins the world championship,” the company’s website says, linking to a press release about the company’s NIST ranking.
In other words, while the U.S. Department of Commerce has sanctioned these companies for human rights abuses, NIST continues to audit the algorithms that got them sanctioned in the first place.
NIST tests nearly 200 U.S.-based and international companies per year through its facial recognition vendor test (FRVT) audit. Many of those companies then use the results of the tests for marketing purposes.
“[NIST is] considered the industry standard and users rely on NIST’s benchmark for their business decisions and purchases,” Shuang Wu, a Yitu research scientist told Wired in March 2019. “Both Chinese and international customers ask about it.”
Although NIST’s algorithm audit provides a valuable service by acting as a neutral third party that vets the accuracy of high-stakes algorithms, it also puts the agency in a unique position. The U.S. government has now found itself…