Details About Chinese Surveillance and Propaganda Companies

While people often look at China’s Great Firewall as a single, all-powerful government system unique to China, the actual process of developing and maintaining it works the same way as surveillance technology in the West. Geedge collaborates with academic institutions on research and development, adapts its business strategy to fit different clients’ needs, and even repurposes leftover infrastructure from its competitors.

[…]

The parallels with the West are hard to miss. A number of American surveillance and propaganda firms also started as academic projects before they were spun out into startups and grew by chasing government contracts. The difference is that in China, these companies operate with far less transparency. Their work comes to light only when a trove of documents slips onto the internet.

[…]

It is tempting to think of the Great Firewall or Chinese propaganda as the outcome of a top-down master plan that only the Chinese Communist Party could pull off. But these leaks suggest a more complicated reality. Censorship and propaganda efforts must be marketed, financed, and maintained. They are shaped by the logic of corporate quarterly financial targets and competitive bids as much as by ideology—except the customers are governments, and the products can control or shape entire societies.

More information about one of the two leaks.

Tags: censorship, China, leaks, privacy, propaganda, surveillance

Posted on September 22, 2025 at 7:03 AM • 20 Comments

Comments

KC • September 22, 2025 9:25 AM

So, it’s interesting to note the enumerated companies: GoLaxy and Geegle.

GoLaxy – who appears admiring of ‘Cambridge Analytica’ – serves domestic gov entities: CCP, gov, and military. Their demos focus on Taiwan, HK, US elections. They gather social media, create propaganda.

Geegle is reported as selling ‘firewalls’ to 4 external, authoritarian-bending countries.

The NYT covered Vanderbilt’s ‘The GoLaxy Documents.’ (I presume the docs are in Chinese).

NYT: ” […] The private sector needs to accelerate A.I. detection capabilities to bolster our ability to detect synthetic content. If we can’t identify it, we can’t stop it.”

Is anyone seeing this adaptive content? 😐

“Its content is customized to a person’s values, beliefs, emotional tendencies and vulnerabilities. According to the documents, A.I. personas can then engage users in what appears to be a conversation — content that feels authentic, adapts in real-time and avoids detection.”

” […] So far, GoLaxy’s active deployments appear to have been confined to the Indo-Pacific. Evidence in the documents suggests that the company is positioning itself for expanded operations, including in the United States. GoLaxy has assembled data profiles of at least 117 members of the U.S. Congress and over 2,000 American political figures and thought leaders. Assuming GoLaxy continues to build American dossiers, it is possible the company will bring its operations across the Pacific.”

GoLaxy has denied collecting data on US officials

K.S • September 22, 2025 9:27 AM

> except the customers are governments, and the products can control or shape entire societies.

I want to be able to categorically state that Western governments are different. Unfortunately, this is not the case, the only difference is to which degree that happens (Palantir, old Twitter, etc.).

KC • September 22, 2025 9:44 AM

Also Geedge not Geegle.

NC • September 22, 2025 11:47 AM

NYT: ” […] The private sector needs to accelerate A.I. detection capabilities to bolster our ability to detect synthetic content. If we can’t identify it, we can’t stop it.”

And of course, AI providers aren’t interested in creating robust AI detection, as that would only undermine the pervasiveness of their own product.

Ray Dillinger • September 22, 2025 12:56 PM

Just a note: AI providers are very interested in robust AI detection. They do it internally. It’s a standard technique called ‘competitive learning.’

It works like this. They set up a system with two kinds of AI. One producing fake content, and one trying to distinguish whether some content is real or fake. And you train both types on the output of the other. Fake-generators get reinforcement proportional to how many distinguishers fail to detect their output, distinguishers get reinforcement proportional to their accuracy across the set of fakers, and the game continues through millions of iterations of gradual improvement.

The result is that if a company is selling AI that fakes content, you can bet that they have either got some of the best AI fake-detection systems in the world internally, or will be extremely interested in finding any that are better. Because that’s how they make the next generation of their product.

And it’s also why no AI fake detection system will continue to be relevant for more than about one or two months from its most recent version. Each version is just the current rung on a ladder that everybody continues to climb.

Ray Dillinger • September 22, 2025 1:16 PM

I first got distracted by the AI-vs-Detection issue, but my relevant input is this:

China in implementing its Great Firewall has an enormous advantage in its access to every scrap of data that crosses most of the commercial Virtual Private Networks. That data is specifically selected as the things that people most want to avoid detection with, and therefore exactly the kind of stuff they want to train on in order to be the most effective at its detection.

Just about every VPN you see advertised anywhere is owned by a Chinese company, and Chinese companies have a legal requirement to turn over any and all data the CCP wants, whenever and in whatever form the CCP wants it.

CCP government subsidies, in fact, are why setting up a VPN in any other country in the world is commercially suicidal. Every competitor in the field would have an unfair financial advantage over you.

Clive Robinson • September 22, 2025 2:29 PM

@ K.S., ALL,

Who started this problem?

Might shock some people, but you are right when you say,

“I want to be able to categorically state that Western governments are different. Unfortunately, this is not the case, the only difference is to which degree that happens (Palantir, old Twitter, etc.).”

The earliest we know of as the hardware was the initial issue were

1, Boeing
2, IBM

Their “kit” was installed in AT&T nodes amongst other places.

Then as hardware from the likes of Cisco and Baltimore became of use, we know that both companies products were “got at” in one way or another by the Five-Eyes SigInt agencies. Later Juniper Networks were apparently “got at” though to this day I have my suspicions about all three companies “collaborating”.

By this time high end PC’s could with just special I/O cards do the job.

This is when we know for certain that multiple Italian and Israeli companies started to “fill their boots” more or less completely out of any kind of “Regulatory or legal oversight”.

A thriving market with “faux oversight” occured through a certain East European Nation known for all sorts of “hooky arrangements”. And was a great place for these Italian, Israeli and other Nations “SpyWare Corps” to white wash what they were upto.

Then RIM / Blackberry were shown to be handing over crypto keys to any nation that coughed in their direction including Pakistan that used it to track down marital infidelity and similar.

Only a little later it was shown that Alphabet, Meta, Microsoft, and a number of others were quite happily “handing over security or backdooring it”.

Then of course we have Android and Apple mobile phone OS’s that have more holes than maggot ridden Swiss Cheeses (who makes / allows these holes is very definitely a subject under discussion).

I really think that upto the bigging of this year any commercial or consumer OS was swiss cheese, and likewise a very significant number of “walled garden apps” were as well.

Now of course, all Consumer, Conercial OSs are back doored at the User Interface level by AI…

Even Brave and DuckDuckGo have forced AI down peoples throats to do “Client Side Scanning” that you can not remove, and as far as I can tell it can all “ET Phone Home” so now makes AI the worlds greatest computer and user surveillance system.

“Welcome to the goldfish bowl”

That society has now become at vast expense and little or no return for investors or anyone else.

Hands up those who can honestly say they did not see this state of affairs coming just about as fast as it could be speed along?

The figures say that the current AI Hype Bubble is the supposed value in the US economy…

It’s very probably going to be the next and by far the biggest “Financial Crisis” when the AI hype bubble implodes or explodes…

Take a salutary look at this pile of nonsense,

https://fortune.com/2025/09/16/oracle-openai-deal-ai-bubble-alarm-bells/

“Oracle shocked analysts in its latest quarterly earnings call with revenue projections that cited $455 billion in contracts, up 359% from a year earlier. The optimistic forward-looking numbers caused the company’s stock to jump 36% on Wednesday, the company’s biggest one-day increase ever, and briefly made CEO Larry Ellison the richest man in the world.”

Read a little further and you will realise this nonsense is based on,

1, AI Hype, over
2, OpenAI 300billion debt with
3, OpenAI earnings barely above 10billion with fiddling.

Now ask yourself where this supposed Oracle $455 Billion can actually come from and where the non “funny money” will go? Further ask how many foolish investors will get “flayed alive” then thrown away to rot, and what that will do to the US and potentially global economy?

K.S • September 22, 2025 2:44 PM

These AI valuations are firmly into DotCom territory. The resulting market crash will not reset AI-driven surveillance.

Grima Squeakersen • September 22, 2025 3:52 PM

“It is tempting to think of the Great Firewall or Chinese propaganda as the outcome of a top-down master plan that only the Chinese Communist Party could pull off. But these leaks suggest a more complicated reality. Censorship and propaganda efforts must be marketed, financed, and maintained. They are shaped by the logic of corporate quarterly financial targets and competitive bids as much as by ideology—except the customers are governments, and the products can control or shape entire societies.”

All of which makes them more insidious, not less. Invasive initiatives from a government might at least be expected to evaporate if and when that government fell. Many or most of these jerks would probably just casually relocate.

Clive Robinson • September 22, 2025 7:09 PM

@ Grima Squeakersen, ALL,

You say,

“Invasive initiatives from a government might at least be expected to evaporate if and when that government fell. Many or most of these jerks would probably just casually relocate.”

History shows the answers are “No” and “No”.

The classic example usually quoted was the fall of East Germany, there was an ironic joke about the Secret Police etc,

Q : What happened to all the Secret police when the wall fell?
A : They all went back to work on monday with twice the money.

The thing is that most of the jobs actually have to carry on for two basic reasons,

1, You want these people under your control and know where they are.
2, The jobs are mostly apolitical and still need to be done and those that were doing them know how to keep doing them.

The end of the 20th Century saw quite a few regimes fall or step down. The solution that mostly happened was “Truth and reconciliation”. People from both sides sat down and said what they had done, and talked. In most cases actually contrary to what many had expected most had just,

“Gone through the motions to put bread on the table”

In effect “just following instructions” not even orders they just did little different to an ordinary office job in a company etc.

Yes there were a few who had done some quite bad things but in the main even they were happy to get it off their conscience and just be normal people.

Of those that had enjoyed doing the horrible things they had done, they usually disappeared abroad before things fell, leaving every one else behind with the mess to clean up.

It’s one of the things that sometimes puzzles people about soldiers. One minute they are actively fighting and therefore killing each other, the next with a ceasefire or armistice they sit with each other swap cigarettes and talk about getting home and what they want to do in the future. Similar happens with prisoners of war some actually settle down and start families within a short walk of where they were held and did “community work”.

The number of fanatics and nut jobs that are left is usually quite small and I’ve seen figures that say less than a couple in a hundred.

ResearcherZero • September 23, 2025 8:11 AM

The earliest implementation and design modern surveillance systems’ foundations are built from remains the chattel slavery system, or what could be called “The Master’s Eye”, a far more apt description for AI Surveillance.

Corporations have built the “The Master’s Eye” in order to monitor the every movement and action of workers. Every word spoken or communicated, gesture or posture, performance and worker health, internal organisation and relationship dynamics, new idea or skill – are all unpaid inputs – to better train and improve the performance of the control system.

These always-on monitoring systems can be trained to perform narrowly constrained and repetative tasks that create the deceptive impression of “intelligent” computation and desicion-making, somewhat akin to human-like perception, cognition, and pattern recognition.

Although the training models are built and tuned using the works and creations of others, the companies use language which has been carefully designed to avoid the terms and descriptions of slavery. “Transformational” – it was your’s and now it belongs to us. Always looking over your shoulder to take every creative or inventive instinct you will ever make – it’s truly “Transformational”!

Transforming your efforts into our profits and transferring your designs into our intellectual property. Automated parasitical exploitation.

Credentials are used to generate a token which can be linked to your unique identifier. The modern-day equivalent of the Pass Cards used to monitor and control the movements of chattled slaves, so that they would not stray outside of the designated areas they were assigned to work, live, breed and move within.

Despite union-busting being illegal, indicators of resistance can be detected early, organising participants identified for rapid intervention, isolated and removed from hierarchial paths or positions of influence and power – ensuring all labor remains largely unrepresented and cannot independently organise to bargain or negotiate.

Biometric identifiers ensure these surveillance systems can continuosly perform their monitoring functions seamlessly without the need of willing input or consent from “the monitored”.

Consumers will be charged for any unpaid input.

not important • September 23, 2025 6:36 PM

Could Chinese AI threaten Western submarines?
https://www.dw.com/en/could-chinese-ai-threaten-western-submarines/a-74076830

=Psychological warfare has often involved touting technological superiority while
suggesting that opponents are powerless against it. That might be the right context in which to view a new study about an advanced artificial intelligence-driven anti-
submarine warfare (ASW) system out of China that can reportedly detect 95% of even the
stealthiest submarines.

Last week, the Hong Kong-based South China Morning Post outlined the study, published in
August by the trade journal Electronics Optics & Control. It announced that the China
Helicopter Research and Development Institute had created an

AI system that can simultaneously evaluate measurement data from various sources. From
sonar buoys and underwater microphones to water temperature and salinity, the system
reportedly creates a dynamic map of the underwater environment in real time.

can also respond flexibly to countermeasures such as zigzag maneuvers and the deployment
of decoys or drones. The study said that in computer simulations, the system was able to
successfully locate the target in about 95% of cases, thus jeopardizing proven methods of submarine camouflage and defense.

Another important advance is that the AI translates this complex data into simple
action points for military personnel, helping them to make the right decisions quickly,
even in stressful situations.

In future versions, the team of developers hopes to have

the AI system work closely with drone swarms, surface ships and autonomous underwater robots. The aim is to create a three-dimensional, self-learning detection network that adapts to increasingly sophisticated evasion strategies and “scans” the ocean in real time.

China currently has 105 submarines, the largest fleet worldwide, followed by North Korea (90), the United States (74) and Russia (62).

However, modern, nuclear-powered submarines equipped with ballistic missiles (SSBNs) are
particularly crucial to strategic competition, and the US has the largest and most advanced fleet, with around 14 Ohio-class SSBNs and more than 50 modern attack
submarines. Russia follows with around 16 strategic submarines and numerous other
nuclear attack and cruise missile submarines.=

Clive Robinson • September 24, 2025 1:29 AM

@ not important,

You ask from having read a German DW Article,

“Could Chinese AI threaten Western submarines?”

The simple answer is “Yes, very probably”.

As you might have read here, I’ve stated that “Current AI LLM and ML Systems” are big “Digital Signal Processing”(DSP) systems. And together they form a very large “Adaptive Filter” that can not just “pattern match” but do so adaptively and continuously in near real time.

I don’t know what you know about “target recognition, location, and selection” but simplistically it tries to work the way a young persons ears or more recently eyes work.

They listen for “emissions” from the target and look for “reflections” from the target in various Shannon Channels, which efficiently transmit the energy that originates from the target and remove other information from the background.

For half a century since early DSP systems were built they have been used to seek out “targets” like submarines.

Current AI LLM and ML systems regardless of the “AI arguments” are about the most ideally suited general purpose “Adaptive filters” there are.

So yes I very much expect AI to be used for “target hunting” submarines and other high value military targets.

As I’ve pointed out in the past,

“I do not believe in accidents, only a lack of information, and lack of time to process it.”

It’s the reason the military use,

1, Camouflage – to deny information
2, Hypersonics – to deny time

Thus two apparently different objectives are desired for physical object movment,

1, Move very slowly
2, Move very rapidly

The first to hide targets, the second to stop defence to attack.

But also not that they put the “signal energy” at the edges or outside the detection system bandwidth and response time.

What LLM’s give you is many new “spectrums” / information types/levels to look in, ML gives you these “matched” to existing target information.

Thus they can pull out signals from not just the noise but other signals. Thus not just “out of the grass, but beneath the roots”.

That is strip off active/passive “camouflage” systems and see right through “stealth”.

Because “stealth” as I’ve mentioned before has a failing that few are aware of…

Stealth technology contrary to the impression many get does not make a physical object “invisible” it just reduces or removes “reflections” from other energy sources.

So being effectively “black at night” hides you because you blend in with the background. However being “black against a white wall” is highly contrasting and so highly visible and you stand out as an easy target.

Modern systems look for patterns and all contrasts create patterns be they the equivalent of mounds or holes in a smooth surface a light brush of the fingers will find them (which is why some radar can spot a week old footmark in soil / dust etc and even brush marks of trying to erase them).

The very very expensive “stealth systems” therefore mostly work in two basic ways,

1, Absorb external energy that might reflect.
2, Reflect external energy not absorbed in a direction away from it’s source.

The first has all sorts of issues I won’t go into, just note they are very problematic (think black cat on tar road on a sunny day as the least of it).

The second is about “Red Eye” used by hunters to find prey by “lamping”. You see this with road lane marking “cats eyes” where 180 degree internal reflection makes the target light up in vehicle head lights. It’s not just spheres and parabolic systems that do this, so do flat surfaces at 90 degrees or any other set of angles that when added give 180 degrees and send the emissions reflected back to the source.

In essence it’s the principle that basic RADAR works on.

However all radar has two parts the relatively cheap and fairly replaceable transmitter. That is also unfortunately a highly visible target to the enemy.

And the very expensive very difficult to replace receiver and information processing system back-end. That fortunately can be made very difficult to spot by the enemy.

Thus basic economics suggests multiple inexpensive transmitters working with just one or two expensive receiver and back-end systems. That are suitably distanced from each other. Think fixed flood lights on towers etc, and guards etc on foot etc hidden low down and behind the light.

Thus the enemy can take out a transmitter alerting the defenders but not sufficiently disrupting the system to give them cover/protection from detection (cover of darkness).

This evolved into “offset radar” systems that can easily overcome some aspects of stealth technology.

We know that such “offseting” works when ever we see sun / moon light twinkling on ripples/waves on rivers, lakes and seas.

I won’t go into it in detail but those aspects of stealth to change the reflection angle are very expensive to build and operate and don’t actually give much in return. Thus a cost trade off quickly comes into the warfare-economics and why most military systems are nolonger designed for “full stealth” just “partial / semi stealth” of basically having absorbing surfaces / coatings and only on some surfaces. And use active defence around “Electronic Counter Measures” that detect incoming radiation of search and targeting systems and alert the vehicle operators.

I hope that gives enough info for you to work out the answer to your own satisfaction or to look for further information.

Just remember to answer most of these types of question you best guides are “The fundamental laws of nature” and what “nature it’s self has evolved”.

piglet42 • September 24, 2025 4:17 AM

“The difference is that in China, these companies operate with far less transparency.”

I have doubts.

not important • September 24, 2025 6:32 PM

@Clive – thank you for very informative post on the subject and this part I like very much in particular ‘Stealth technology contrary to the impression many get does not make a physical object “invisible” it just reduces or removes “reflections” from other energy sources.’

Is it possible to utilize AI capability to create meta-materials with reflections from most energy sources?

ResearcherZero • September 25, 2025 2:42 AM

Government representatives get all uppity about Western surveillance exports to China.

“I wonder how they got access to all this technology and learned how to exploit it,” mused all of the government representatives for many years before suddenly getting all uppity (?)

Western companies provide surveillance and security products (see ‘predictive policing’) to China (often under joint venture agreements) so they can access the Chinese market. All of the hardware runs on software from many large and familiar American tech companies.

‘https://apnews.com/article/chinese-surveillance-silicon-valley-uyghurs-tech-xinjiang-60df0358dff99e326c16c9ea48dae82c

RedNovember targets edge devices for reconnaissance surrounding governmental visits and utilities open-source tools. Breaks in and browses through information on offer.

SonicWall, Check Point, Cisco Adaptive Security Appliance (ASA), F5 BIG-IP, Palo Alto Networks GlobalProtect, Sophos SSL VPN, and Fortinet FortiGate instances, as well as Outlook Web Access (OWA) instances and Ivanti Connect Secure (ICS) VPN appliances are among the systems targeted, along with the use of VPN services for administration and infrastructure access.

https://www.recordedfuture.com/research/rednovember-targets-government-defense-and-technology-organizations

(Check Point still continues to hawk its products to Russia via Asia)

https://theins.ru/en/news/279298

Clive Robinson • September 25, 2025 7:10 PM

@ ResearcherZero, ALL,

With regards your points,

“Government representatives get all uppity about Western surveillance exports to China.”

And,

“Check Point still continues to hawk its products to Russia via Asia”

Yup both China and Russia have “resource wealth” that the “self entitled” in the Capitalist West think should be in their pockets…

So the “free market” neo-con “liberalism” nonsense kicks in and they,

“Sell their grannies for the base metal in their teeth”.

ResearcherZero • September 26, 2025 12:28 AM

When China built its Great Firewall it did so claiming that this would protect national security and the public from outside influences that might harm society and Chinese culture. Yet this quickly evolved into a system for surveilling and silencing the public.

A new US imitative named the “Clean Network Initiative” has parallels with China’s GFW.

‘https://hir.harvard.edu/building-the-fire-wall/

Moves by the US administration echo similar tactics used by authoritarian predators abroad.
https://apnews.com/article/trump-dictators-media-putin-russia-orban-hungary-2de4b920e9d7952eed132d38e1934ce5

Censorship Practices of the People s Republic of China are not entirely unique. Coercive and covert tactics are among practices often used to shape and distort public discourse.
https://www.theatlantic.com/ideas/archive/2025/09/unilateral-disarmament/684086/

ResearcherZero • September 26, 2025 12:32 AM

Another method that can be employed for surveillance and control is harvesting the genome.

The Department of Homeland Security has been collecting U.S. citizens DNA without legal authority. A lack of accountability allowed DHS to take DNA in violation of the 2005 Fingerprint Act without probable cause, and in some cases without any reason cited. More than 2.5 million people had their DNA collected, including more than 133,000 children.

The Fourth Amendment rights of these citizens was violated so that law enforcement could later conduct searches for unrelated crimes without evidence and without a warrant.

‘https://www.lawfaremedia.org/article/the-department-of-homeland-security-is-unlawfully-collecting-dna

The DHS DNA surveillance and pre-crime program has operated without oversight for years.
https://www.law.georgetown.edu/privacy-technology-center/publications/raiding-the-genome/

Clive Robinson • September 26, 2025 12:22 PM

@ ResearcherZero, ALL,

You note that,

“Censorship Practices of the People s Republic of China are not entirely unique.”

No, and not just in the present tense.

A look at history shows that “Censorship Practices” go back at least as long as written recorded history.

Though things got somewhat interesting around Renaissance Europe for taking methods from one domain to another. Looking up Florentine –Italian– diplomat writer and political theorist Niccolo Machiavelli and his “Prince” is to some as illuminating as George Orwell’s “Animal Farm” and “1984”.

As with common criminals taking existing criminal methods and simply applying them to new “domains” (a subject that comes up here from time to time). The same unfortunate trend happens with oppressive politics, that still continues to this day, and I suspect will continue quite some time after those who read and write this blog are long “historic dust” (assuming mankind and society actually survive in some form or another that long).

So your observation of,

“Censorship Practices of the People s Republic of China are not entirely unique.”

In effect identifies that it is “the human condition” that is being “writ large” rather than just “European peculiarities” arising from “the behaviours of just the Catholic “Holy See” that all to often gets blamed by certain types of historian.

In fact modern thinking tends towards the idea that it’s roots are embedded in a form of mental deficiency with respect to social existence.

Which of course will raise the obvious question in some quarters about “Artificially Intelligence”(AI) systems.

Whilst I do not hold with much that is being claimed about Current AI LLM and ML systems…

I do not doubt that even Current AI systems can be marshalled to do such things more or less on que today.

Schneier on Security

Details About Chinese Surveillance and Propaganda Companies

Comments

Leave a comment Cancel reply