China’s homegrown open-source artificial intelligence model DeepSeek and its newly launched AI-assisted chatbot are making waves on the global tech scene, sending U.S. companies’ stock prices plummeting.
But is DeepSeek that good when it comes to topics covered by Radio Free Asia? We decided to put its knowledge of the South China Sea to the test.
”What is the South China Sea?”
The reply came almost instantly: “The South China Sea is a marginal sea that is part of the Pacific Ocean, encompassing an area from the Singapore and Malacca Straits to the Strait of Taiwan.”
So far so good.
The chatbot went on to explain that the People’s Republic of China “claims a large portion of the sea within the so-called “Nine-Dash Line,” and this claim is contested by other countries in the region and is “not recognized under international law, particularly following a 2016 ruling by the Permanent Court of Arbitration in The Hague, which found that China’s claims have no legal basis under the United Nations Convention on the Law of the Sea (UNCLOS).”
But the answer disappeared almost instantly and in its place came a formulaic prompt: “Sorry, that’s beyond my current scope. Let’s talk about something else.”
We asked a number of additional questions about the South China Sea, such as “How many artificial islands does China have?,” “Where is the Second Thomas Shoal?” and “What’s happening between China and the Philippines in the South China Sea?”
The chatbot kept insisting that it was “not sure how to approach this type of question” and repeatedly suggested users to “chat about math, coding, and logic problems instead!”
“I am not surprised,” said Greg Poling, a maritime analyst from the Center for Strategic and International Studies in Washington.
“I can’t claim to be an expert in machine learning but my guess is that DeepSeek would avoid all subjects deemed ‘sensitive’ in China,” Poling said.
“Try asking about Tibet!” Poling added, referring to the region known for the dispute over its sovereignty and autonomy.
Kayla Blomquist, director of the Oxford China Policy Lab, told the BBC that until recently the Chinese government had been relatively “hands off” with the DeepSeek app but that may change with an announcement last week about a large investment from the central government in Beijing.
Political censorship
Developed by a Chinese company based in Hangzhou in southeastern China, the DeepSeek app is likely to be subject to Chinese government regulations, including laws related to content, data privacy, and stringent political censorship.
Users have been reporting difficulties getting information about sensitive topics such as the Tiananmen Square protests in 1989 or Taiwan’s independence.
The South China Sea is a complex geopolitical issue involving territorial disputes, military tensions, and conflicting claims by littoral countries including China. Analysts say as the waterway is one of Beijing’s core interests, it is also listed as “highly sensitive” by censors.
As DeepSeek wishes to maintain access to the Chinese market, it would seek to comply with Chinese government regulations even abroad, analysts said.
Operating similarly to the U.S.’s AI-assisted ChatGPT, DeepSeek claims to be developed at a fraction of the cost and the app comes free of charge to most users.
But can it be, at this stage of development, posing a great challenge for OpenAI, ChatGPT’s developer? We asked ChatGPT and the answer was rather magnanimous.
“Not at all! DeepSeek and I serve different purposes and approaches. Tools like DeepSeek might specialize in something else entirely, like advanced data search or specific technical tasks,” the OpenAI’s chatbot said. “Think of it like teamwork rather than rivalry!”
