“Where has this mosque gone?” tweeted a University of British Columbia law student on Tuesday, referring to the Keriya Aitika Mosque in China’s Xinjiang Region. The centuries-old building “disappeared in early 2018 despite [being] a Major Historical and Cultural Site Protected at National Level,” meaning destruction would require State Council approval.
The destruction of mosques has been reported before in Xinjiang, and online investigation site Bellingcat picked up the thread this time around. “By using open sources and satellite imagery,” they explained, “we can locate the mosques and check the claims. We can also potentially narrow down when the alleged destruction took place.” Bellingcat used mapping applications to confirm the “loss of beautiful and historic buildings, such as the gatehouse at Keriya Mosque and a large portion of the Kargilik Mosque.”
On Wednesday, a bipartisan group of U.S. lawmakers publicly urged the U.S. government to do more on Xinjiang, where state repression of the Muslim population includes detention camps and blanket surveillance. And whilst ‘doing more’ should include sanctions against key Chinese officials, the politicians also demanded that U.S. companies stop collaborating with China’s surveillance state machine and that U.S. financial institutions stop funding Chinese technology companies, including AI unicorns, whose systems power Xinjiang. The investment case for those start-ups is the surety of programs funded by Beijing. There are no excuses.
“Americans would likely be very troubled, if not outraged,” the politicians pointed out, “to learn that their retirement and other investment dollars are funding Chinese companies with links to the Chinese government’s security apparatus and malevolent behavior—links that represent material, asymmetric risks to corporate reputation and share value.”
And then on Thursday, the New York Times published undercover imagery from the citywide prison. The team reported on police checkpoints where Muslims are photographed before being sent on their way, on compulsory communications monitoring and citizen scoring, and on the discouragement of worship, where mosques are equipped with high-tech surveillance and kindergarten children are “interrogated” about the religious activities of their parents.
Just another week in the life of China’s dystopian surveillance state.
The AI arms race
The world now seems inured to the repression of Xinjiang’s Muslim population. Despite media reports and political rhetoric, there is nothing new in these surveillance claims. And yet Beijing continues on regardless. Nothing that has been done by the international community has had any kind of meaningful impact.
The letter to the U.S. Secretaries of State, Commerce and Treasury, urged the government “to expand its ‘Entity List’ to include businesses and other entities that have provided technology, training, or equipment to Xinjiang officials in mass detentions and surveillance, including Chinese companies like Hikvision and Dahua Technology.”
U.S. sanctions and prohibitions against Hikvision and Dahua are now in place. And the same is true for the communications giants Huawei and ZTE that develop the core infrastructure. We ask that you “ensure that U.S. companies are not assisting, directly or indirectly, in creating the vast civilian surveillance or big-data predictive policing systems being used in Xinjiang.”
Last week, the Massachusetts Institute of Technology bowed to U.S. pressure and announced that it would cease collaborating with Huawei (and ZTE). But what about SenseTime? Last year, MIT announced a joint effort to “advance artificial intelligence research.” At the time, SenseTime was a major shareholder in SenseNets. You’ll remember SenseNets from the data breach earlier this year that exposed a database of more than 2.5 million sets of personal details and 6.5 million records relating to the GPS locations (recorded by facial recognition) passed by those individuals in the prior 24 hours. There was no immediate comment from MIT on the ongoing relationship with SenseTime at the time of publishing.
SenseTime sold down their 49% stake in SenseNets before news of the data breach became public, and they have insisted that they did not supply the facial recognition technology being used. After their exit, SenseNets became fully owned by its other major shareholder, Netposa, a Chinese company with offices in Silicon Valley and Boston, and which has received and made U.S. investments.
SenseNets also claimed a partnership with Microsoft (denied by the U.S. company), and the breach appeared to show Azure Cognitive Services sitting within the surveillance solution. Victor Gevers, the hacker responsible for publishing the breach, tweeted at the time that, “the company 微软 also known as Microsoft has been a precious partner who has turned more than once a blind eye to the (technical) / (mal)practices of the engineers of SenseNets.”
When I asked Microsoft about the alleged SenseNets partnership and the use of Azure Cognitive Services, they said that they are not “involved in a partnership with SenseNets,” adding that, “we’ve done a search of all our partnerships over the past five years and don’t have any evidence of Microsoft having a partnership with SenseNets. We’ve done a thorough examination. SenseNets and its parent company are not customers of our Azure services including those related to facial recognition, and we have no evidence they’ve purchased our products or services in the past five years.”
Actions or more words?
This week, those 24 senators and 19 representatives asked whether “American companies [have] sold technology to Chinese companies operating in Xinjiang or [have] worked with them to help develop the ‘Sharp Eyes’ or ‘Integrated Joint Operations Platform’ being used there for surveillance and big-data predictive policing? And what specific steps are being taken by the U.S. Government to prevent the Chinese government from using American-made goods and services to perpetrate grave human rights abuses in Xinjiang?” The truth is that such links are extensive.
SenseTime raised $620 million last year, in a funding round that confirmed its status as “the world’s most valuable unicorn in artificial intelligence” with a capitalization of more than $4.5 billion. “The funding was led by Fidelity International, Hopu Capital, Silver Lake and Tiger Global and followed a $600 million funding round in April led by Alibaba Group.”
On its website, SenseTime references “more than 700 customers and partners in China and overseas, including world-renowned institutions and companies such as the Massachusetts Institute of Technology (MIT), Qualcomm, NVIDIA, Honda, Alibaba, Suning, China Mobile, UnionPay, Wanda, Huawei, Xiaomi, OPPO, vivo, Weibo.”
This is not an isolated example. The California State Teachers’ Retirement System and the New York State Teachers’ Retirement System remain investors in Hikvision, despite the company having been sanctioned over Xinjiang. Hikvision’s revenue growth slowed last year, but they still managed to grow in each quarter, to top $7 billion for the year. Surveillance states make for lucrative investments.
In buying China’s state-subsidized electronics and participating in these investment rounds, the West is an apparently willing participant in this. If software from these heavily state-influenced ‘start-ups’ is used to oppress populations towards the ‘total surveillance’ dream, thus leading to increases in revenue and capitalizations, then it becomes a frightening case of being careful what you wish for.
In the world of surveillance, if there are no impediments applied by technologists, customers, investors or regulators, if AI development continues to accelerate fueled by these investments and partnerships, then how does sitting on the sidelines criticizing the destruction of religious buildings or the oppression of a religious population come across as anything but hypocrisy.
We are now in April 2019. Does anyone think the situation in Xinjiang will be any different this time next year?