AI chatbot DeepSeek celebrated in China despite concerns elsewhere

When Ye Shengwei first tried using China's new artificial intelligence (AI) chatbot, DeepSeek, to write articles, he was surprised by its high quality. "It provided in-depth analysis and explained why it generated that content," said Mr. Ye. "It provided unexpectedly good perspectives on the topics I raised and drew some useful conclusions." Ye Shengwei was impressed by DeepSeek's capabilities.

Mr. Ye, 35, a software product manager at a large technology company, had previously used Western AI chatbots to draft work articles. Due to the Chinese government's "firewall" restrictions, ordinary citizens cannot access Western websites. Mr. Ye stated that he used to think ChatGPT was the best AI tool, but after using a domestically produced Chinese chatbot, he might change his preference. "ChatGPT's output still requires me to manually edit and polish it sentence by sentence, which is still time-consuming, and it tends to give vague conclusions," Mr. Ye said. "If DeepSeek can improve my writing efficiency, I will continue to use it."

Since the release of its new AI model R1 on January 20th, DeepSeek has been making global headlines, shaking up the stock market, amazing computer scientists, and inspiring pride in China. Its creators, High-Flyer Quant, a previously little-known AI hedge fund company, claim that the model was built at a much lower cost than AI models like ChatGPT, and is more efficient, amid US restrictions on the latest technology. DeepSeek's AI assistant—a direct competitor to ChatGPT—has become the number one free application downloaded on the Apple App Store, with some worrying that the Chinese startup has disrupted the US market. Its release coincided with Donald Trump's return to the White House, with Trump calling it a "wake-up call" for US companies. Chinese netizens have shared their excitement on social media, with one netizen describing it as the "best Lunar New Year gift." One Chinese social media user posted: "I think DeepSeek is fully capable of the job, and the logic is clearer, very suitable for Chinese students." "I have unsubscribed from ChatGPT and plan to stick with DeepSeek."

While the Chinese government, official media, and users have all praised DeepSeek, others have expressed concerns about censorship and potential surveillance by Beijing. Built-in restrictions limit DeepSeek's responses to questions involving sensitive topics in China, such as the 1989 Tiananmen Square incident. Some, including an Australian cabinet minister, have also warned that people should be cautious when using DeepSeek and pay attention to their data security. Housing Minister Clare O'Neil said on Channel Seven's "Sunrise" program: "It's okay to talk to the application, but maybe don't give it personal information that you don't want the rest of the world to know." Science Minister Ed Husic told ABC News that there are still many unanswered questions, including "data and privacy management." He added: "I would be very careful. These types of issues need careful consideration."

Nicholas Davis, an AI expert at the University of Technology Sydney, said that many platforms censor content, and in China, "the government regulates what is displayed." He said: "All AI models are fine-tuned to move the output in a specific direction." According to Professor Davis, scientists have successfully removed censorship from some open-source Chinese models, and in theory, DeepSeek could also do this. "An important question is, in the case of DeepSeek, are people worried about specific censorship topics? Or are people concerned that it is a Chinese company?" Professor Davis stated that data security is more concerning than censorship, and that this is relevant to all AI platforms. "Personally, the bigger issue is the risk of entering personal information into [online large language models (LLM)]," he said. "For example, I would never enter any sensitive work information into a [internet] browser-based large language model because I know that companies like Anthropic and OpenAI can see and potentially use what I write. The US is considered to be a leader in AI, and the export bans implemented are intended to maintain this position. China's new AI tools are challenging these assumptions."

Professor Davis said: "I have the same concerns about DeepSeek. If I worked for the government, I might be more worried about using DeepSeek." "It's not about censorship. It's about 'what happens to my data?' and 'how might that data be used?'" "But, if everyone only uses DeepSeek, then, yes, censorship is a critical social issue, because freedom of speech supports democracy." "But I think privacy is the bigger issue right now." "It's easy for people to think, 'I'll use this AI to organize my finances.'" "For any LLM, this exposes sensitive information that could come back to hurt you."

Professor Davis said that he has been testing and benchmarking DeepSeek, which can be installed on a user's computer for local use, rather than through the internet. He said it is "very good" and he is impressed by its efficiency. "If I'm on a long flight, I can now run a very powerful model directly on my MacBook during the flight," he said. "I don't need an internet connection, nor do I need a server to run ChatGPT or [Anthropic's AI chatbot] Claude for me. I can do it locally." He said that despite concerns, open-source models like DeepSeek can help people better understand the technology more broadly. "If everything is completely controlled and managed by large tech companies, it means there are fewer opportunities for independent experts and communities to participate, test, and understand it," he said. "DeepSeek is a very interesting real breakthrough, and we should be deeply involved in it."