Dahua Responds, Caught Lying
By John Honovich, Published Feb 11, 2021, 11:51am ESTDahua has responded to IPVM and the LA Times investigation into Dahua's "Uyghur Warnings". Dahua has been caught lying repeatedly about its human rights violations.
Dahua admitted that the Dahua documents IPVM discovered are authentic:
Based on the Company’s internal review, the relevant documents reported by certain media are historical internal software design documents.
The 'history' for these documents is very recent - March 2019 to December 2020, just 2 months ago.
In December 2020, Dahua's 'big data' guide for its PRC police projects listed support for tracking "Uyghurs with hidden terrorist inclination' alongside drug addicts, thieves, prostitutes, fugitives, gangsters, fraudsters, excerpted below:
This was an internal document that Dahua thought would never be made public.
Lied To SCMP
However, a month before that, in November 2020, Dahua told the SCMP:
Dahua Technology does not sell products that feature [an] ethnicity-focused recognition function [emphasis added]
Xinjiang vs Rest of PRC "Warnings"
Now, Dahua's new statement, following the LA Times / IPVM report, does not deny selling ethnicity detection, just not in particular 'regional markets':
does not provide products and services for ethnicity detection in such regional markets. [emphasis added]
Dahua's Director of Marketing Jiaqi Gao emailed IPVM today qualifying that the region is Xinjiang:
We never provided products or services for ethnicity detection in Xinjiang, PRC
But Uyghur detection in the Xinjiang Uyghur Autonomous Region is not the point, since Xinjiang is the Uyghur homeland, and such warnings would go off constantly. Indeed, the PRC has much more draconian ways of controlling Uyghurs in Xinjiang.
The point of Dahua selling and operating "real-time Uyghur detection" is to alert police throughout the rest of China when a Uyghur is spotted outside of Xinjiang. It is big business for Dahua and terrifying for Uyghurs who are warned against, using Dahua video surveillance, for the 'crime' of being Uyghur.
Dahua lying to the SCMP was not their first time lying about persecuting Uyghurs.
Lied After Sanctions
Dahua posted a similar 'notice' after the US sanctioned Dahua for human rights abuses of Uyghurs in October 2019, declaring:
With regard to U.S. Department of Commerce’s decision on adding Dahua Technology to the Entity List, we express our strong protest to such decision, which is in lack of any factual basis, and call on the U.S. government to reconsider on it. [Emphasis Added]
The factual basis was significant including Dahua constructing and operating police stations in Xinjiang.
Moreover, our new investigation found a March 2019 Dahua test report of its face analytics mentioning "real time Uyghur warnings" thirteen separate times, including:
As can be seen above, Dahua marked this as "Pass", meaning "real time Uyghur warnings" was included/functional 7 months before US sanctions were enacted.
Not In The Future
Dahua's new statement pledges that they will not offer this 'in the future':
Dahua will not provide the features or applications in the software products in the future.
Dahua has repeatedly lied about its actions yet now Dahua expects the public to believe they will not do it again.
Dahua has created an alarming situation with its combination of systematic human rights abuses and flagrant lying to the public.
4 reports cite this report:
Comments (39)
Dahua will not provide the features or applications in the software products in the future.
I think they forgot something.
Dahua will not provide the features or applications in the software products in the future in such regional markets.
Better.
Just saw this on Baidu AI site (which is an equivalent to our AI.Google) bragging about their pretrained classifiers that any solution vendor can hook up to products. Some of these would make us think twice about the ethical aspect, but for the Chinese it's quite OK. Different culture!
Well, they 'misstated' (lied), got caught, and feel they have constructed an acceptable reason for being (how do they state it in the US and elsewhere.....'less then truthful'...no thats not it....I know we can control the language with one or more of these...
fabricate, fib, prevaricate, perjure, equivocate, fudge, deceive, delude, dupe, fool, gull, hoodwink, snow, trick, falsify, misreport, misrepresent, misstate, distort, misinform, mislead
Yes I was lazy - the above terms were borrowed from the pages of Merriam Webster ). Maybe they should have just come up with a code-word or number to describe the ethnicity (among other analytics) in their code and documentation. They really should have had a bean-bag sit-down with the folks at G***l* and Amaz** (the G-word and A-word) - now, they know how to profile everything. Better document control and less feature description(the 'just figure it out for yourself' method) in the future, may be tried, as the demand will most likely not simply go away. My gosh, how are we (the rest of the world) to re-train them (and others) not to profile by ethnicity or any other data-set so the need no longer exists - guess is we won't. Where there is a need, greed, and ability, someone will fulfill the need. I wonder how the UK, US and AU are so good at concealing (or just getting away with it) the fact that they ('they' being a variety of LE agencies, non-LE agencies, public and private companies, even private individuals) too are profiling everything under the sun via ever evolving AI analytics. Maybe just force those that want the 'bad' capabilities to develop the custom code themselves (wait..some already do). Maybe only order products with NO AI analytics (educate customers on the potential misuse scenarios, but really, in that deep, dark part of their minds they don't talk about, they still want it...all of it) and if a MFG doesn't offer products AI-free, then watch as profit margins plunge; bring them to their knees. But wait, they may not need 'regular commercial' customers with govmt funded/sponsored-clientele . Maybe we shouldn't pick and choose which analytics are 'good' or 'bad'; they may all eventually be bad (they may already be..depends on who you're speaking with). None of our customers really want to know about suspicious activities, known-thieves, prostitutes, drug-users/dealers, particular ethnicity presence(what does ethnicity have to do with anything anyways - there isn't any statistical data supporting the need for this), known child molesters or any of that silly data/meta-data (even if historical statistical data supports bad things happening as relate to the various data-sets) - don't lie to yourself, yes they want it.. Particular data-set categories and options of said categories are viewed by some to be simply immoral and unethical. Some will also argue that humans are so much better at determining who is OK and who isn't anyways, since the beginning of time. Maybe, force those lazy PRC to be on the lookout with other humans for anything or anyone they feel they need to monitor, torture or kill in the name of 'natl security'. Some may say shame on Dahua (and others) for inventing solutions to support diabolic needs; no one has ever done that before or will again (who else comes to mind........). But remember; AI and analytics are Grrrreat for everything else. Good job for outing those bad guys; liars, in a time where lying has become the 'norm' (that was not sarcasm, truly); hopefully you have much much much more time on your hands to get the rest (of the bad guys). It may be time to begin training your kids, kids-kids and kids-kids-kids to become investigative and unbiased reporters; we probably need more of those.
Nice job reporting, as usual.
-Cheers from the realm of reality.
What else are they doing as this "wasn't supposed to be made public"?
Is anyone surprised? You cant believe anything they say, only what they do. The west has pretended this is not the case with China for decades because of financial incentives. When will the west wake up? We have plenty of ways we could play hardball with them but choose not to.
Wondering (as usual) who chooses the images and motion GIF's associated with the articles ( this one and other Uyghur-related in particular)? Maybe I simply misunderstand but I was under the impression the reporting here was 'unbiased' and 'impartial'.... I'm sensing potential bias with the choice of imagery and as we all have been told one time or another, " a picture is worth a thousand words". What are these pictures actually saying? The images used in some of the articles may be construed as to attempt to evoke certain, possibly even negativly biased, emotions before even reading.. Luckily, I'm not subject to respond emotionally to this imagery, but some may certainly fall prey to it. Maybe I'm just imagining it.... not questioning the articles' other, actual written content here; the content is usually pretty informative and at times even entertaining.
Thank you for your unbiased reporting.
Im pretty sure this clip will illustrate how the US will respond to Duaha
@john there is no question on why this is unethical. Did IPVM research/test the technology on how the analytics work? Is it based on skin color? Would love to read more about that

02/14/21 06:35pm
IMO China, The Motherland, made Dahua and then financed Dahua to create the needed s/w needed to pick out races. In China's leaderships minds, it furthers the success and directives of China's goals. Dahua is likely doing as told or their CEO may take some mandatory time off. Kinda like the CEO of Alibaba et al. Dahua has one important customer all the rest can be forgotten. I just don't see the business need for race ID and I bet Dahua did not either.
"Don't learn the tricks of the trade, learn the trade."
