DeepSeek storm: More collaboration or protection?
This article is authored by Sriparna Pathak.
The arrival of DeepSeek in January 2025, took the world by storm, and DeepSeek found mentions in not just the Artificial Intelligence (AI) world, but also in related realms of international relations and geopolitics. Based in Hangzhou of the Zhejiang province of China, DeepSeek is owned by the Chinese hedge fund High-Flyer. The Chinese company develops large language models (LLMs). In contrast with OpenAI, which is propriety technology, DeepSeek is open source and free, and challenges the revenue model of United States (US) companies that charge monthly fees for AI services. What caused the storm in the AI world was that DeepSeek disrupted the prevailing notion that the development of LLMs needs significant technical and financial resources. However, DeepSeek released its LLM at a fraction of the cost that other vendors incurred in their own developments. As a result of this, investors began casting doubts on the value of large AI vendors from the US, including Microsoft, Meta, Nvidia, and Oracle among others. All witnessed significant drops as investors reassessed AI valuations. However, Elon Musk and Palmer Luckey among others cast doubts on the Chinese startup’s claimed budget and chip use.
In a post on X, Luckey stated that DeepSeek is pushed by a Chinese hedge fund to slow investments in American AI startups, “service their own shorts against American titans like Nvidia, and hide sanction evasion”. Similarly, Alexandr Wang, the CEO of Scale AI, cast doubt on DeepSeek, stating that it had “had access to 50,000 more advanced H100 chips that it could not talk about due to US export controls”. In December last year, the Biden administration made it harder for China to build its own AI chips as it announced a sweeping set of export controls on chip-manufacturing equipment, memory and software, which made it more complex for Chinese companies like Huawei and ByteDance to develop cutting-edge AI.
A doubtful budget for creating the LLM is least of DeepSeek’s problems now, which has already been the target of a significant cyber-attack, leaked API secrets, publicly accessible database information, chat history and back-end data, and sensitive user information. Security researchers have also raised concerns about DeepSeek’s neglect of elementary security measures. More cyber-attacks are not difficult to predict, as any widely used online platform becomes a constant target for hackers. Ivan Tsarynny, CEO of Feroot Security, an Ontario-based cybersecurity firm focused on customer data protection, stated in an interview that direct links to servers and to companies in China that are under the control of the Chinese government are not difficult to see, and this is something that has never been seen in the past.
Beyond data, DeepSeek also seeks to normalise China’s occupation of other countries’ territories, and a direct link between the Chinese state narrative and the narrative churned out by DeepSeek is not difficult to find. For example, when asked to answer questions on India’s Arunachal Pradesh, the chatbot simply refuses to answer the question, stating that it is beyond its scope and asks to chat about something else. The situation is no different when the chatbot is asked prickly questions about Xi Jinping, human rights repressions, Taiwan, and China’s 1962 war against India. Clearly, there are unanswered questions about the chatbot’s programming, potential biases, and its implications on the future of AI interactions. While DeepSeek holds the top spot in global downloads, the list of countries where it is downloaded the most is headed by India. Given that China engages in state-sponsored disinformation against India, DeepSeek’s responses on Indian territories claimed by China is an attempt to normalise the discourse and the learning on China’s territorial aggression. If the LLM were to be used for the next few years by the young generation, China’s narratives on the territories it claims from other countries, including India’s would be normalised.
As such, some countries have banned DeepSeek with different extents. Australia banned DeepSeek from all government devices and systems over the security risk it poses. The US navy also has banned DeepSeek, owing to potential security and ethical concerns. New York also moved to ban DeepSeek, citing foreign surveillance and censorship concerns and the harvesting of user data. South Korea has also banned DeepSeek, as regulators have stated that they are suspending the app until the can be sure that it complies with the country’s data protection laws. Taiwan’s digital ministry, has also urged its government departments to not use DeepSeek. India, which is still locked in a conflict with China, despite disengagement in just two areas should take a note from other countries’ stances. While cooperation on tech and AI are great for ensuring a stable international order, allowing AI and tech with malicious motives and inadequate data protection can only mean furthering of conflicts. As such, the usage of DeepSeek needs to be more cautiously undertaken, if at all.
This article is authored by Sriparna Pathak, associate professor, Chinese Studies and International Relations, Jindal School of International Affairs, OP Jindal Global University, Sonipat.