November 16, 2019

AI Weekly: In China, you can no longer buy a smartphone without a face scan

Last Updated on

Picture a world in which the baseline requirement for a new smartphone is a facial recognition test. You needn’t imagine it — in China, beginning December 1, that’s the scrutiny to which the country’s over 850 million internet users will be subject, without exception. The Ministry of Electronics and Information Technology said in a late September statement that the rule, which will also prevent people from exchanging their mobile numbers, was necessitated by an uptick in phone and internet fraud. Furthermore, the ministry pledged that it would help to “increase supervision and inspection” and “strengthen assessment accountability” while “supervis[ing] the implementation of work” and the “real-name registration management of telephone users.”

Tying ID verification to phone purchases isn’t exactly unprecedented. A rule imposed by the German interior ministry in 2016 requires that domestic telecoms ask customers for ID cards, foreign passports, or temporary ID papers when they buy a SIM card or a cell phone. Under French law, carriers must collect identifying information of all users and subscribers of prepaid services. And in the U.S. in 2010, New York Sen. Chuck Schumer (D-NY) and Texas Sen. John Cornyn (R-TX) introduced a bill that would have mandated ID verification for every prepaid phone purchase, ostensibly to deter terror suspects.

But China’s new regulation — a harsher imposition than the registration system that launched in 2013, which requires customers obtaining a new phone number to volunteer their IDs — is among the first with a facial recognition component. Despite an outcry on Chinese social media, it seems unlikely that critics will prove successful in pushing against it, for the reason that sales provide valuable data to cross-reference with internet activity. (See: China’s sprawling social credit system.) Shipments hit 97.6 million in the second quarter of 2019 alone.

It’s an open secret that China is compiling the world’s most extensive facial recognition database — one with the power to identify any one of its more than 1.3 billion citizens within seconds. (The core corpus is said to be between 13 terabytes and 90 terabytes in size.) A long-term goal of the project, which the Ministry of Public Security launched in 2015 in conjunction with Shanghai-based security company Isvision, is to match faces to ID photos with 90% accuracy. Mum’s the word with respect to the Ministry’s ulterior motives, but perhaps tellingly, cameras equipped with Isvision’s technology in Tiananmen Square were connected as early as 2003 to police databases of suspected criminals.

Complicating matters, public opinion is far from universally against the new regulation. That’s perhaps a symptom of facial recognition’s skyrocketing adoption on the mainland, where it’s become more or less commonplace. Alibaba’s Alipay division this year committed 3 billion yuan (roughly $421 million) to spur retail sales of its upgraded Smile-to-Pay system, which enables customers with linked accounts to pay for goods almost solely with their faces. Xiaozhu, a Beijing-based booking website for daily rental and short-term rooms, announced last December plans to add facial recognition-enabled door locks to 80% of its listings in Chengdu. And two years ago, KFC teamed up with tech giant Baidu to predict customers’ orders by their estimated age and mood, as inferred from their facial characteristics.

But the new buying requirement is uncomfortably juxtaposed with China’s ongoing AI-aided surveillance efforts, most recently spotlighted by a ban in Hong Kong of masks and face paint that aims to deter protesters from evading the city’s face-recognizing cameras. Colleges like Peking University track students with systems installed on campus gates. Facial recognition technology was implemented at two border checkpoints between Hong Kong and Shenzhen in a crackdown on resellers. Cameras capable of facial ID have been deployed at concerts to catch criminals. And special glasses with facial recognition software have been distributed to police in select cities like Zhengzhou, where they were used to search for suspects at a high-speed rail station.

China’s facial recognition industrial complex was the recent target of the U.S. Commerce Department, which blacklisted SenseTime, Megvii, Yitu, Dahua, Hikvision, and other leading AI surveillance startups in the country including for enabling “mass arbitrary detention” of citizens, as well as the surveillance of Uighurs, Kazakhs, and other members of Muslim minority groups in Xinjiang. But that seems unlikely to discourage the Chinese government, which has plenty of incentive — and no real disincentive domestically — to continue expanding its use of facial recognition into countless domains. And given both its history of human rights abuses and facial recognition systems’ propensity for prejudice, the outcome looks dimmer by the day.

For AI coverage, send news tips to Khari Johnson and Kyle Wiggers — and be sure to subscribe to the AI Weekly newsletter and bookmark our AI Channel.

Thanks for reading,

Kyle Wiggers

Senior AI Staff Writer