In the past year, several Chinese companies have released powerful language models—DeepSeek, Qwen, Hunyuan, Ernie—that are not only technically impressive but, more importantly, entirely open-source. These are not token gestures. The models are complete. The code is public. The weights are downloadable. The licences are permissive. Anyone can use them—for free. This is an extraordinary development. It is not just a triumph of Chinese ingenuity. It is certainly not to the benefit of the ageing bureaucrats who want to make the entire Chinese population into caged and monitored featherless bipeds. Artificial intelligence is now leaving the hands of the few, and entering the hands of the many.

When we say a model is “open-source,” we mean several things—each of them significant.

1. Source code is published and readable. Anyone can see how the model works under the hood—its architecture, training processes, tokenisation logic, and pre/post-processing layers.
2. Model weights are released. These are the numerical parameters that define the model’s knowledge and behaviour. Without them, you can’t run the model. Open-source means you can.
3. Licensing permits free use and modification. Models like DeepSeek-V2 or Qwen are typically released under permissive licences—Apache 2.0, MIT, or BSD-3. These allow not only academic or personal use but also commercial deployment, provided you give appropriate credit and don’t sue the creators.
4. Fine-tuning is allowed. You can take the base model and retrain it on your own data, creating specialised tools for law, medicine, engineering, translation, or customer service.
5. Redistribution is permitted. You can share your improved model with others. This creates a self-reinforcing loop: each new contribution benefits the global community.

By contrast, proprietary models like OpenAI’s GPT-4 or Anthropic’s Claude do not permit any of this. Their source code is closed. Their weights are secret. You can only access them through APIs that are tightly controlled, rate-limited, and monitored. You are a user, not an owner. Open-source AI restores ownership.

Artificial intelligence is not just another gadget or software trend. It is a general-purpose technology—like electricity, the printing press, or the internet. It can be applied across fields, layered into existing systems, and used to accelerate other kinds of discovery. Its potential to improve the human condition is almost unmatched. Consider just a few examples:

• Education: AI tutors can teach children in any language, adapt to different learning speeds, and provide one-on-one help in subjects where human teachers are stretched thin. In a school with 30 pupils and one teacher, AI can ensure no child is left behind.
• Medicine: AI models trained on vast medical datasets can help diagnose diseases earlier, suggest treatments based on individual genetics, and interpret imaging scans with precision. They don’t replace doctors, but they make doctors vastly more effective.
• Law and administration: AI can parse complex regulations, draft documents, detect fraud, and speed up bureaucracy. It can ensure that benefits reach those entitled to them—and expose those exploiting loopholes.
• Scientific research: Already, language models are being used to summarise academic literature, identify overlooked connections between fields, and even propose new hypotheses. As their capabilities grow, they will become indispensable lab assistants.

These are not futuristic dreams. They are already happening in pilot form. But their reach depends entirely on access. If AI remains locked up behind billion-dollar paywalls, the benefits will flow only to the wealthiest corporations. Open-source models change that. They allow universities, charities, small companies, and individuals to build locally relevant tools without waiting for permission.

China, by opening its models to the world, has broken the monopoly. The question now is who will make use of this opportunity—and who will squander it.

Britain has missed every major wave of computing since the 1970s. We failed to capitalise on early microprocessors. We surrendered the browser and the search engine to America. We never built a major mobile phone OS or a serious cloud platform. Instead, we drifted into dependence.

Now, though, the field is open again. A university in the North of England or a startup in Cardiff can download DeepSeek-V2, fine-tune it on British legal precedent or NHS guidelines, and deploy it on secure local servers. A school in Newcastle can build a maths tutor that speaks with a regional accent. A Welsh charity can train a bilingual assistant for elderly care. All without needing Silicon Valley.

But to do this, the government must stop obstructing and start enabling. The costs of electricity in the UK are among the highest in the world, largely due to Net Zero policies that prioritise virtue signalling over viability. Running even modest AI workloads becomes a luxury. Taxes on capital investment remain punishing, especially for firms that haven’t yet turned a profit. Worse still, the legal environment is increasingly hostile to free expression. The Online Safety Act has created a culture of self-censorship among developers. No one wants to build a tool that might be labelled “harmful” by a civil servant.

We don’t need a complete industrial revolution. We need five or six sensible changes:

• Suspend Net Zero requirements on data centres and research labs.
• Cut VAT and corporation tax for AI startups under a certain revenue threshold.
• Create exemptions from speech laws for academic and technological research.
• Allow local councils and schools to procure AI tools directly, without centralised approval.
• Make the public sector default to open-source models when available.

These aren’t radical measures. They are common sense. They could be enacted this year. And they would make Britain a serious participant in the AI age, not just a supplicant to American firms or a commentator on Chinese ones.

One of the most encouraging effects of the Chinese model releases is the emergence of truly global development. Until now, cutting-edge AI lived in three places: California, Beijing, and maybe Zurich. Everyone else was a consumer.

But now, teams in Brazil are building fine-tuned chatbots for court systems. Researchers in Egypt are adapting models for Arabic legal texts. Developers in rural Vietnam are using open weights to create crop-planning apps based on weather data and soil conditions. There are open forks of Chinese models in Basque, Swahili, Farsi, and Scots Gaelic. We are witnessing a decentralisation of intelligence. AI is no longer something imported. It is being made locally, with local voices and local knowledge.

This is why openness matters more than ever. If the models are free and modifiable, they can evolve to fit every part of the human world. If they remain closed, we will be forced to live with whatever assumptions, biases, and limitations the original designers chose to bake in.

Britain, which prides itself on independent thinking, should instinctively prefer the former.

Of course, none of this implies that China has become an apostle of freedom. Its internal censorship is severe. Its surveillance state is real. Many of its companies operate under state guidance. Some of its models—especially those trained on Chinese-only data—include political filters that no Western liberal could accept. But DeepSeek, Qwen, and others are trained on global corpora, published with English documentation, and released with full weights. There is no backdoor. No usage limit. No geo-lock. In technical terms, they are genuinely open. And that’s what matters. If the West had done the same, I would be writing about that instead.

Artificial intelligence will not wait. Whether it’s disease modelling, defence logistics, school planning, or pension administration, these systems will be integrated into every part of national life within a decade.

The only question is who controls them—and who gets left behind. China has thrown open the gates. That may have been a calculated move, but it was a gift all the same. As said, it may not be to the long-term benefit of such ambitions as the Chinese Government may have. It is now up to countries like Britain to decide: do we join the builders, or do we remain the customers?

The tools are there. They are free. They are legal. And they are ours to use.