“What the AI” is happening in China?

Dearest Gentle Readers, the ton is still quite bamboozled with all the AI shenanigans that have refused to slow down in any way. I sincerely apologise, if you’re not a Bridgerton fan you won’t have a clue of what I’m talking about. But if you are, then it’s simply a case of - IYKYK.

I’m still trying to figure out which is more interesting - Bridgerton or AI!

So let’s get into the juicy tidbits of AI and what’s happened across the world in the past couple of weeks. Before I get started on that, how many of you, like me, are wondering how this will all end up? Because it’s still quite obvious that AI is a cash cow and hasn’t yielded reasonable ROI compared to the high levels of investment it has received over the years.

Well, while you’re mulling about the future - not just the future of work but the future of humanity - the AI Labs are still building faster and seemingly more capable models, while the rivalry between 2 of the world’s largest economies - the US and China, fires up.

Can you guess who’s topped the chart this past week on the most capable multimodal models? It’s China. China seems to be going ahead of the US in the AI race. I mean it’s quite obvious that the geopolitical AI race is only between the 2 countries - China and the United States. 

Here’s a snapshot of some of the latest AI developments you need to know about: 

• DeepSeek V-4:  China’s Deepseek released DeepSeek-V4 last week, a 1.6-Trillion-Parameter Mixture-of-Experts model, which simply means, a large frontier model. It’s multimodal with code and text capabilities and surpasses the performance of the world’s most advanced closed-source models like GPT-5.5 and Claude Opus 4.7. It costs one-sixth of the cost of America’s closed-sourced models.

It’s also open-source and available under a commercially-friendly MIT licence, meaning anyone can use, modify, sell, and distribute software for free using the model.

• ChatGPT Images 2.0: ChatGPT images 2.0 features new thinking capabilities that allows it to actively search the web for context, before reasoning through the structure of the image. It supports various languages such as Japanese, Korean, Chinese, Hindi and Bengali in addition to English.

This language expansion gives a loud hint of OpenAI’s market expansion plans. 

• Amazon Quick: AWS launched Amazon Quick, a mix of an AI Agent and personal AI Assistant that is hosted in a local environment. It runs continuously in the background, connecting all fragmented enterprise tools into a single proactive AI layer. It connects apps like Slack, Gmail, Zoom, Salesforce and Microsoft 365. Amazon Quick creates presentations and interactive dashboards.

• Claude Design: Anthropic isn’t holding back with the recent launch of Claude Design, an AI visual creation tool that generates interactive websites, UI prototypes, dashboards and presentations directly from text prompts, images or documents. It provides an Agentic UI where you describe, edit and export designs. 

The AI lab also formed a partnership with Adobe and launched new creative connectors that allows Claude to connect directly with Adobe Creative Cloud. The connectors also work with other design apps such as Affinity, Blender and Ableton. 

• Mistral AI: The french AI start-up, founded by my ex-colleagues from Google Research, recently released Mistral Medium 3.5, a new 128-billion parameter dense model with a 256K context window. This basically means there is another very large frontier model with lots and lots of compute. In addition, they also launched “Work Mode” for Le Chat, their multilingual conversational AI assistant, taking autonomous coding agents from local environments straight into the cloud. 

• Nvidia: The major AI chips supplier launched Nemotron 3 Nano Omni, a 30 billion parameter reasoning open-weight multimodal model designed to process text, images, video and audio in a single architecture. It uses a hybrid mixture-of-experts (MoE) architecture that works with vision and audio encoders integrated directly into the system, which allows the model to deliver 9x faster throughput than all other open omni models. 

Let’s also take a look at what’s happening in the AI law, regulation, and privacy space.

As the geopolitical AI wars take steam, the White house announced measures to counter the "unauthorised, industrial-scale distillation” of American AI models by foreign actors, specifically Chinese firms. The administration is exploring legal measures to hold identified actors accountable. OpenAI and Anthropic have accused China on several occasions of using distillation methods (a process of using the output of an AI model to train large models which mimic the original model). 

The UK government woke up one bright sunny morning and found the NHS’ (The UK’s National Health Service) data on sale on Alibaba - China’s largest marketplace, also known as the “Amazon of China”.  The data belongs to 500,000 volunteers participating in UK Biobank, a major health research project. While the data was de-indentified which means it didn’t include names, addresses, or phone numbers, it had information such as gender, age, date of birth, socio-economic status, lifestyle habits, and health outcomes which if combined with other data, would make it easy to identify the affected individuals. This issue was reported as a data leakage rather than a cyberattack.

My head is literally spinning with all these update, new model releases and apps, global wars, data theft and so on. 

The biggest question remains - how does all of this affect the average person working daily to pay their bills, and live a happy and satisfying life?

What does the introduction of the latest AI designers, sorry agents mean? What should graphic designers and people working in the creative industry do? Here’s a simple tip: embrace the changes, work with your favourite AI apps,  and most importantly, maintain your human client relationships the old fashioned way. You know, like take them out for dinner, check in on your clients, have parties, be your most charming self - do all the things those AI agents can’t do! 

As a creative, or small business owner, or marketing professional, if you were already hooked on Midjourney, then got lured to ChatGPT for image generation, then swayed to Google’s nano-banana, and now have a plethora of AI design apps such as Claude Design, which has now been integrated into Adobe Creative Suite, you’re probably beginning to wrack up a massive migraine. These never-ending launches and competition definitely puts a strain on brand and customer loyalty. 

So we might end up like the large OTAs such as Booking.com and Hotels.com who had to come up with creative ideas to drive customer loyalty. I remember my days working as the EMEA Product Lead for Google Travel, and the conversations around customer loyalty and competition. The competitive pressure led Booking to introduce its genius loyalty programme to ensure customers stay loyal and happy. I wonder if we’ll soon be seeing some form of loyalty programmes coming up soon in the AI industry. 

There’s also a lot going on in the regulatory law landscape with recent breaches identified by the Digital Services Act from the European Commission, accusing Meta of breaching their own terms and conditions and allowing children and minors below the age of 13 to use Instagram and Facebook. 

In the same vein, the Chinese government recently issued interim measures for AI Anthropomorphic interactive services, a major step in regulating AI powered emotional services such as companionship, emotional care and support. 

It’s the very first global regulation specifically addressing the issues of anthropormorphisation which means the act of giving human qualities, emotions or behaviours to non-human, inanimate objects - such as AI. In this context, it also refers to treating or imagining AI as if it were human or sentient - which it’s definitely not.

China’s new law sets the precedent for protecting children against the deception and misleading behaviours from AI companions and chatbots. Considering children and minors have been the most negatively impacted by these, they’ve taken the lead on ensuring that some of the dangers of AI are swiftly addressed before further harm is done. A great step in the right direction.

But to sum this up, China has taken the lead on developing extremely cheap and highly capable AI models, while stealing and selling people’s data on their global marketplace and at the same time releasing the first important law on child protection against the dangers of AI companions and chatbots. This is definitely a mixed-bag, similar to our current reality.

If you’re already using DeepSeek or thinking about switching to the cheaper Chinese models, it’s important to point out that data from Chinese AI companies is owned by the Chinese government. This means once you use DeepSeek, your data belongs to the Chinese government. 

And if your head is spinning like mine on the best AI tool to use for your purposes, take this simple advice, save yourself the headache and stick to what works, while investigating and trying out the new toys - sorry apps on the sidelines.