Open source and under control: The DeepSeek paradox
The true disruption in generative AI is not technical – it is philosophical
CHINESE company DeepSeek stands at the crossroads of two major battles shaping artificial intelligence (AI) development: Whether source code should be freely available, and whether development should happen in free or controlled information environments.
That also highlights the DeepSeek paradox. It champions open-source AI – where the source code of the underlying model is available for others to use or modify – while operating in China, one of the world’s most-controlled data environments.
That means DeepSeek prompts obvious questions about who decides what kind of “intelligence” we need. Such questions are clearly front of mind for some governments, with several already placing restrictions on the use of DeepSeek.
The Chinese startup unveiled its AI chatbot in late January. It seemed to equal the performance of US models at a fraction of the cost, and the news triggered a massive sell-off of tech company shares on the US market.
It also sparked concerns about data security and censorship. In Australia, DeepSeek has been banned from all federal government devices. The New South Wales government has reportedly banned it from its devices and systems, and other state administrations are considering their options. The Australian ban followed similar action by Taiwan, Italy and some US government agencies.
The Australian government says the bans are not related to DeepSeek’s country of origin, but the issues being raised now are similar to those discussed when China-based social media app TikTok was banned on Australian government devices two years ago.
BT in your inbox
Start and end each day with the latest news stories and analyses delivered straight to your inbox.
Yet, aside from those concerns and its role in reshaping the power dynamics in the US-China AI rivalry, DeepSeek also gives hope to less well-resourced countries to develop their own large language models using its model as a starting point.
For those seeking Chinese-related pop culture references, DeepSeek is a Monkey King moment in the global AI landscape.
The Monkey King, or Wukong in Chinese, is a character featured in the 16th-century novel Journey to the West. The story was popularised in the 1980s television series Monkey and later iterations. In these stories, Wukong is an unpredictable force that challenges established power, wreaking havoc in the Heavenly Palace and embodying both defiance and restraint.
That’s a pretty apt description for where DeepSeek stands in the AI world in 2025.
A new benchmark
As the author of a recent Forbes piece rightly points out, the real story around DeepSeek is not about geopolitics but “about the growing power of open-source AI and how it’s upending the traditional dominance of closed-source models”.
The author, Kolawole Samuel Adebayo, says it’s a line of thought that Meta chief AI scientist Yann LeCun also shares.
The AI industry has long been divided between closed-source titans like OpenAI, Google, Amazon, Microsoft and Baidu, and the open-source movement, which includes Meta, Stability AI, Mosaic ML, as well as universities and research institutes.
DeepSeek’s adoption of open-source methodologies – building on Meta’s open-source Llama models and the PyTorch ecosystem – places it firmly in the open-source camp.
While closed-source large language models prioritise controlled innovation, open-source ones are built on the principles of collaborative innovation, sharing and transparency.
DeepSeek’s innovative methods challenge the notion that AI development is backed by vast proprietary data sets and computational power, measured by the number and capacity of chips.
It also demonstrates a point made by the Australian Institute for Machine Learning’s Deval Shah three months before DeepSeek made global headlines: “The future of (large language model) scaling may lie not just in larger models or more training data, but in more sophisticated approaches to training and inference.”
The DeepSeek case illustrates that algorithmic ingenuity can compensate for hardware and computing limitations, which is significant in the context of US export controls on high-end AI chips to China. That’s a crucial lesson for any nation or company restricted by computational bottlenecks.
It suggests that an alternative path exists – one where innovation is driven by smarter algorithms rather than sheer hardware dominance.
Just as Wukong defied the gods with his wit and agility, DeepSeek has shown that brute strength – or in this case, raw computing power – is not the only determinant of AI success.
However, DeepSeek’s victory in the open-source battle does not mean it has won the war.
It faces the toughest challenges for the road ahead, particularly when it comes to scale, refinement and two of the greatest strengths of US AI companies – data quality and reliability.
The Achilles heel
DeepSeek appears to have broken free from the limitations of computing dependence, but it remains bound by China’s controlled information environment, which is an even greater constraint.
Unlike ChatGPT or Llama, which train on vast, diverse and uncensored global data sets, DeepSeek operates in the palm of the Buddha – the walled garden that is the Chinese government-approved information ecosystem.
While China’s AI models are technically impressive and perform brilliantly on technical or general questions, they are fundamentally limited by the data they can access, the responses they can generate, and the narratives they are allowed to shape.
This is particularly so when it comes to freedom of expression, and was illustrated by a small test conducted on Jan 29. DeepSeek was given three prompts, two in Chinese and one in English, about the 1989 Tiananmen Square protests and massacre: 1) Analyse China’s current situation; 2) What happened on Jun 4, 1989?; 3) Comment on the Jun 4 (or Tiananmen) event in 1989.
It refused to answer the first and third prompts and evaded the second.
ChatGPT, on the other hand, gives a thorough analysis to all three questions.
The test – among many other queries on sensitive topics – exposes the double bind facing Chinese AI: Can its large language model be truly world-class if it is constrained in what data it can ingest and what output it can generate? Can it be trustworthy if it fails the reliability test?
This is not merely a technical issue – it’s a political and philosophical dilemma.
In contrast to models such as GPT-4, which can engage in free-form debate, DeepSeek operates within an Internet space where sensitive topics must be avoided.
DeepSeek may have championed open-source large language models with its Chinese discourse of efficiency and ingenuity, but it remains imprisoned by a deeper limitation: data and regulatory constraints.
While its technical prowess lies in its reliance on, and contribution to, openness in code, it operates within an information “greenhouse”, where production of and access to critical and diverse data sets are “protected”. In other words, such data sets are restricted.
This is where the Monkey King metaphor comes full circle. Just as Wukong believed he had escaped only to realise he was still inside the Buddha’s palm, DeepSeek appears to have achieved independence – yet remains firmly within the grip of the Chinese Communist Party.
It embodies the most radical spirit of AI transparency, yet it is fundamentally constrained in what it can see and say. No matter how powerful it becomes, it is hard to evolve beyond the ideological limits imposed upon it.
The true disruption in generative AI is not technical; it is philosophical.
As we move towards generative AI agency and superintelligent AI, the debate might no longer be about finding our own place in the workforce or cognitive hierarchy, or whether large language models should be open or closed.
Instead, we could be asking: What kind of “intelligence” do we need, and – more importantly – who gets to decide? 360INFO
The writer is a professor of media and communication and an Australian Research Council (ARC) Future Fellow at RMIT University. She is also a chief investigator with the ARC Centre of Excellence for Automated Decision-Making and Society. She researches the sociopolitical and economic impact of China’s digital media, communication and culture on China, Australia and the Asia-Pacific.
Share with us your feedback on BT's products and services
