DeepSeek: AI model prompts international tensions

By Zara Hossain

The release of DeepSeek-R1 by the Hangzhou-based company High-Flyer has left investors on Wall Street shaken – $600bn of Nvidia’s market cap was wiped off the markets virtually overnight, sinking 17%. ChatGPT was at the forefront in the streamlining of AI, until DeepSeek’s bombshell revelation. 

AI chatbots are powered by Large-Language Models (LLMs) that can generate human-like conversations and responses to user prompts. There are a few things which set DeepSeek apart from its other Western chatbot counterparts, primarily that it is low-cost and has made its model open-source. Virtually anyone can access the AI at little to no cost, evening out the tech playing field and reducing the barrier to entry. This creates a somewhat democratising effect on AI by widening the potential for accessibility. Companies can switch to DeepSeek as a low-cost alternative with little to no difference in the product output, driving more competition and investment in business ventures. Increasing the efficiency of AI can also address the environmental concerns over the use of AI; it reduces its carbon footprint by reducing the processing energy needed. So why has DeepSeek caused such a ruckus? 

American companies have kept their training methods under lock and key to monopolise their AI innovation

American companies have kept their training methods under lock and key to monopolise their AI innovation, and have spent over $100bn in investments in AI companies. The US has also actively undermined the development and accessibility of AI through its restrictions on chip exports to China. Despite this, DeepSeek costs less than $6m worth of computing power from Nvidia H800 chips, a modified version of the banned H100 chips that were stockpiled before export restrictions. Restricting technology and competition has proved to make China more innovative in the face of new obstacles. The US continues to create more barriers to competition while China is opening up technology, highlighting America’s prioritisation of maintaining a firm monopoly on its AI IP to stay ahead in the new age ‘AI-war’, leveraging criticism towards OpenAI and its CEO, Sam Altman, for its secretive and extractive model. Even if DeepSeek is not better than ChatGPT in its capabilities, it has revealed the costly and inefficient nature of American AI innovation, prompting investors to look East. Trump has asserted this to be a ‘wake-up call’ for American big tech, reminiscent of the Cold War ‘Sputnik crisis’ in which Western nations were anxious about the technological gap between the US and the Soviet Union. 

The AI war is a reflection of wider geopolitical tensions

The AI war is a reflection of wider geopolitical tensions. China’s technological innovation is invariably perceived as a security issue which undermines the Silicon Valley tech monopoly. Western countries have responded to the new chatbot with suspicion, accusing it of data misappropriation – many have banned or initiated investigations into the chatbot. Problems remain with DeepSeek; concerns have been raised over its censorship and non-responsiveness. Some users took to asking about the 1989 Tiananmen Square massacre and the issue of Taiwan to highlight the chatbot’s state censorship and unreliable responses. Ed Husic, the Australian Science Minister, has warned against the use of DeepSeek due to concerns over its privacy policy, even though Gemini, ChatGPT, and other AI chatbots and social media platforms collect similar volumes of data from users. DeepSeek has been accused by OpenAI of using unauthorised data to train its model, despite the multitude of IP violation lawsuits which have been directed at OpenAI. 

Other models in the market include Qwen by Alibaba which claims to outpace DeepSeek. New models emerging may and could potentially allow countries outside the US and China to become more cutting-edge, and may present a wider trend toward increased transparency in the AI sphere.

Image credit: Jakub Halun via Wikimedia Commons