The Trap Mechanism: By offering “free” foundational models, China aims to make the Global South algorithmically dependent on Chinese infrastructure, embedding censorship and surveillance capabilities into foreign digital architectures.

How is this different from what OpenAI, Claude and others are doing? They’re offering “free” use of models, losing mountains of money, all in the hopes of becoming the sole names responsible for the infrastructure governing other countries.

Let’s be fair, most big media outlets do little to no reporting when it comes to AI, they just take any and all PR pieces at face value, so this site doing the same, or being a paid mouthpiece, wouldn’t be much of a stretch.

Meanwhile, US tech giants like Microsoft, Google, and Amazon, which expected high-margin returns on their multi-billion-dollar AI investments, face a pricing collapse

This here shows how out of touch these companies are with their own fucking customers. How the fuck did they expect to burn 200+ billion dollars and get any ROI from AI? The demand is not going to grow into infinity and, contrary to what Sam Conman and others say, people would rather dump AI than pay per token usage when they can’t even tell how many tokens they’re burning. Even without China in the mix, they would crash and burn.

Once these nations build their digital infrastructure, government services, and private sectors on these models, they become dependent on Chinese cloud infrastructure, updates, and regulatory frameworks, compromising their digital sovereignty.

This is true IF AND ONLY IF said countries decide to pay for Chinese cloud services. Which, in case anyone is wondering, is deciding to pay, dunno, AliCloud instead of AWS. It has nothing to do with AI.

Oh, and fun fact, many governments (nation, state and city) around the world have some data in USA datacenters, but USA good, so they definitely don’t spy or steal any of that, for realsies!!!

Was this article sponsored by OpenAI and Anthropic?

If a nation or company purchases inference hardware to run open source chinese AI, the article makes it out to be that they are locked in.

With open source china will undercut American companies

Personal opinion alert: GOOD!!!

But that’s the neat part about owning your own hardware you can switch to any model you want at any time.

Along with fine tunes so that you can get chinese models to say that taiwan is a country, this really reads like a red scare article to get people to not use open source and instead use proprietary American.

There is a fundamental error in the analysis here: the author doesn’t seem to understand what open source is nor what the cause of AI commoditising AI will be.

Firstly, by it’s very nature open source AI models are very unlikely to be “trojan horses”, because the code is fully available to the countries, companies, and groups taking the models. And as the author even mentions, the allure is taking the technology and training them on your own curated data. To suggest that China’s models are open to surveillance by China just reads like an American political talking point shoe horned into the article to drive fear, and in stating that they seem to be behaving as shills for America’s proprietary, private corporation driven AI push.

China is certainly behaving strategically by making it’s models low cost and freely available. They’re not trying to steal the worlds data (that mindset really is very Silicon Valley - that’s the model for all the big tech companies after all), but they are trying to standardise the globe onto an open standards compliant version of AI they have a hand in designing, because that will benefit them economically. And it’ll likely benefit the whole world barring the US and it’s tech giants. Why? Because then no one is locked out of the new technology via inflated costs and intellectual property claims of a few American companies. China don’t need the data and they don’t need to spy on the rest of the world; instead they want to shape the entire market so they can compete and build on it. That’s not to say China is trustworthy, nor that actors in China won’t abuse this. But the fundamental drive is not “surveillance”; it is economic & strategic. The global response to that need to be to work with China and ensure the standards satisfy everyone, not just China.

Secondly China’s behaviour on it’s own is not going to commoditize AI - it’s actually a likely outcome for AI that is going to happen anyway. The US AI bubble is predicated on a major assumption: One winner will take all; someone will make the best biggest model and be the Google or Microsoft or Facebook of AI. That’s why huge sums of money are being thrown at scaling up AI data centres to have the most powerful model. But for that to work financially, AI has to be proprietary and locked away; they need to be walled gardens that you have to pay to access (or hand over your data) or you miss out. And there is a built in acceptance that the “losers” in this race are going to have wasted their money. It’s a gamble that they’re backing the winners in this winner takes all world. A stock market bubble is unsurprising with this mentality. It’s not being replicated in other countries, including China, where there is a lot of AI research going on in both public and private institutions. This is a very American take on AI.

And that assumption is based on a flawed economic idea: the idea people will go to one big AI tool in the same way they went to one search engine, and that dominance will be unbeatable. They assume people will pay premium prices for this amazing model or be locked in to this one model as it’s too good to compete with. In reality people are cost conscious - they will pay for cheapest tool that does the job they want. So in AI - why pay a premium price for a model that is vastly over powered for the job you want to do? Instead people will pay for the cheapest model that can do the job well enough - and that may often be free.

AI by it’s very nature can easily be commoditized in this way. People can already run very capable LLM and image diffusion AI models on their own PC if they want to, but also anyone with some money can set up an AI server in the cloud to do basic tasks. There is a huge range of opportunities to compete with a big AI company.

That wasn’t the case with search (it’s taken years for Google’s share in search to start to decline) or social media (the biggest network attracts more people because it IS the biggest). But with AI, while the biggest engine is impressive, users just don’t need it to write an email or tag their face in their photos or generate an image of a bear at a board meeting.

The US AI industry’s response would be that the models will be free for many uses, and it’s the advertising and user data that will turn a profit, much like it has with search and social media. But that is based on the assumption that the costs will come down with scale (at the moment the exact opposite is happening) and also another very flawed assumption: that the US AI models will be a global rich world winner. Their own “Risk assessment” makes the assumption clear:

High Risk: Global bifurcation of AI standards. The West operates on high-cost, secure, proprietary models, while the Global South relies on free, Chinese-subsidized, surveillance-capable models. This creates a severe intelligence and economic asymmetry.

“The West”? No, Europe, Canada, Australia, Japan etc are not locked in to the US or Chinese models; there is an awful lot of research and development going on in a lot of places. And Europe is a good example of how things are changing: They’ve been burnt by Donald Trump so many times they have learnt the US is not a reliable partner any more. They’re not enemies by any stretch, but there is no longer acceptance of unbalanced relationships with the US offset by benefits elsewhere. That means digital sovereignty has become a watch word in strategic planning. OSs, cloud computing and a data platforms, and of course AI. It is foolish to believe Europe is going to just jump in with US AI models, and allow these companies to do with AI what they did with search, social media, and cloud services.

Overall, I think there is a degree of scaremongering over China’s AI models (which matches what the US AI industry and politicans are pushing) that doesn’t match the reality, plus there a mindset from the US AI stockmarket / investment bubble that is drowning out some of the realities of the commoditization of AI and the seismic geopolitical changes that are occuring in the West thanks to the US behaviour in recent years.

TLDR: “China publishes open source AI models - but at what cost???”

None of the arguments in the article make sense, no evidence for the claims is provided, and the language is emotionally manipulative.

This is clearly written from a sinophobic or imperialist mindset, where everything China does is automatically ascribed an evil motive, and everyone of the 1.5 billion people living there is suspect.

China’s aggressive push for open-source AI is not a gesture of digital altruism

Oh it’s now aggression to publish open source software with some kind of look towards benefits from doing so! The language here is so blatantly emotionally manipulative it’s almost funny. The author unironically calls China an “Open-source aggressor”

Hilariously, meanwhile India is “Pragmatic balancer”. What is this lol.

In any case, open source models can be run on other hardware, so the dependence on cheap Chinese infrastructure is temporary at best.

The model will inherently struggle with, or refuse to generate, content related to Taiwanese independence, the Tiananmen Square massacre, or critiques of the Chinese Communist Party

Evidence? The open version of Deepseek has no such restrictions. It’s questionable this is even technically possible.

And who doesn’t know about Tianamen Square? Do they thing Chinese people seriously do not know about the Tianamen Square massacre? People move on and care about other things.

And what right do we have to support Taiwan independence?

An academic should know that these topics constitute agitation as propaganda to foment unrest and instability in geopolitical enemies. US spends billions on this, it’s not even in question. They seek regime change in China, which is an actual act of aggression. That many countries around the world take steps to stop foreign actors influencing their population is not anti-democratic or anti-liberal. In fact it’s anti-liberal to support and whitewash such foreign influence. Chinese people can think and decide for themselves.

PS: Just saw this about western models. So ironically China’s open models that can be run without a harness and specific filters, and can be edited and re-trained, could support freedom of expression much better than western models.

Second, there is the alignment and ideological trap. Open-source models from China are trained to adhere to “socialist core values.”

What DOES make those values evil except that China is evil by default? Can you name them?

I do not think this is technically feasible. If we COULD bake in socialist core values and ideology into an LLM model, that would actually be really great.

It simply assumes that “China bad” or “socialism is evil”. Without argument it’s just “it is known”. Apparently only through neoliberal economics and plutocrats owning the AI models can humanistic values be achieved. Which is patently false. Just compare quality of life and feminist rights and safety and actual social liberty between socialist China and capitalist India. Or point to an academic study that does so.

This fractures the global internet further, creating a “splinternet” where not only the applications differ, but the very cognitive engines processing information operate on divergent ethical and ideological frameworks.

What the heck are cognitive engines with divergent ethical and ideological frameworks? What nonsense is this?

Even if this was true, who in their right mind wants a unipolar world with a single neoliberal capitalist mindset?

Counterarguments: The Case for Open Ecosystems

… technologists argue that the US push for closed, proprietary AI creates a techno-feudal system where only a few billionaires control humanity’s cognitive engine.

While these points hold merit regarding the general value of open-source technology, they fail to account for the specific nature of the Chinese state.

Again, China is so bad, so evil, that even an open source gift of a free AI model must be rejected. The author made a pro forma attempt to show counter arguments only to immediately dismiss them.

But this technology is bad and makes everything it touches worse. Free foot guns for all is not support.