Thread by @CharlesRollet1

new: @AlibabaGroup openly offers Uyghur alerts as a service on @alibaba_cloud, allowing customers to be alerted any time Alibaba detects a Uyghur in video or images. via @ipvideo

here is @ipvideo's full report ipvm.com/reports/alibab…

and here is Alibaba API guide showing Alibaba offering Uyghur detection perma.cc/9ZUV-UD2F

Alibaba deleted it after it was contacted for comment

and read @zhonggg's report in @nytimes about this here nytimes.com/2020/12/16/tec…

Alibaba Cloud's Uyghur detection is part of a content moderation solution, Cloud Shield, which "detects and recognizes text, pictures, videos, and voices containing pornography, politics, violent terrorism, advertisements, and spam"

this means that @alibaba_cloud customers (thanks to AI) can be notified any time a Uyghur shows up in a video/picture.

Hypothetical example:
1) a Uyghur decides to livestream a passionate speech
2) Alibaba AI detects a Uyghur
3) the stream is flagged for review ASAP & removed.

(of course, if toggled on, the Uyghur detection would work for all content, not just politically sensitive speeches but even a prerecorded video or photo album)

@alibaba_cloud claimed, without any evidence, that the API guides with "Uyghur" and "ethnic minority" detection were only for "testing". Alibaba did not explain why it would 'test' Uyghur alerts in the first place.

reminder: $BABA is a NYSE listed firm worth over $700 billion

IPVM also identified Kingsoft Cloud, a NASDAQ-listed PRC cloud provider worth ~$10 billion, as having "Uyghur, non-Uyghur" detection in its API guide, which even detected that someone's face was 48.7% Uyghur (whatever that means)

d1tzzns6d79su2.cloudfront.net/uploads/embedd…

Uyghur-targeting AI is rampant in China's surveillance firms, which sell Uyghur alerts to Chinese police for their security camera networks.

Alibaba's Uyghur rec shows this explicitly racist software is even more pervasive, extending into the country's thriving internet economy

It's worth noting that Kingsoft stated
"labelling on the basis of any race is inappropriate and inconsistent with Kingsoft Cloud’s policies and values" while Alibaba has remained silent on this point.

Huawei and Alibaba, two firms that have admitted to testing Uyghur alerts/detection, are Platinum Sponsors this year of @NeurIPSConf, the world's most prestigious #MachineLearning / AI conference nips.cc/Conferences/20…

the one reported example I could find of face analysis AI being used for China censorship was @JoshuaDummer's experience of getting blocked from a livestream taiwannews.com.tw/en/news/3963271

Alibaba Cloud offers "Asian"/non-Asian face alerts, but unknown what AI did this to Joshua

UPDATE - @AlibabaGroup has a new statement: "Racial or ethnic discrimination or profiling in any form violates Alibaba’s policies and values. We are dismayed to learn that Alibaba Cloud developed a facial recognition technology in a testing environment..." alizila.com/statement-from…

Here was the initial statement by @alibaba_cloud