China's answer to ChatGPT: disruptive and dangerous
This week people in China welcomed the year of the snake. The zodiac animal is known for its intelligence as well as its danger, which feels very relevant given the biggest story that has come out of China this week - DeepSeek.
By now most will be familiar with DeepSeek, the Chinese startup that just launched its latest AI model, one that appears to match OpenAI's capabilities at a much lower cost. It is very intelligent artificial intelligence. But it is also dangerous. Because it operates out of China, even if you’re accessing it elsewhere, you’ll be fed responses from behind the Great Firewall. Looking for straight answers to questions about Tiananmen Square, the persecution of the Uyghurs and the Hong Kong protests? Forget about it.
According to the organisation NewsGuard, DeepSeek’s new chatbot model failed to provide accurate information on news and information topics 83% of the time. That’s a staggering figure and one with huge implications if DeepSeek becomes a global AI leader.
Another danger is to users. DeepSeek says in its privacy policy that personal information it collects is stored in secure servers located in China. Just how secure are these? Let’s see: in 2017 China passed a National Intelligence Law, which compels all Chinese companies to assist the government with national security matters. This means, in theory at least, that any company can be forced to share user data with Chinese authorities, even if the data is from users outside the country. And DeepSeek collects a lot of data, from user’s accounts and activities to data from the devices they’re using and even “keystroke patterns or rhythms,” which can be as uniquely identifying as a fingerprint or facial recognition.
It's not just the Chinese Communist Party though who could access user’s data. Wiz, an organisation that helps other organisations create secure cloud environments, looked into the security of DeepSeek more broadly. Within minutes of investigating they said they discovered a database linked to DeepSeek, which was open and exposing sensitive data.
DeepSeek has emerged on the global stage less than a fortnight after millions in the USA flocked to RedNote, the popular Chinese social media app. This app (which goes by the less than subtle name Xiaohongshu in China, that’s “Little Red Book”) also hoovers up user data and filters conversations along CCP lines.
The popularity of RedNote (and for that matter TikTok) has raised important questions about how Beijing is trying to shape global narratives and the implications for free speech. With the emergence of DeepSeek, answering these questions is even more urgent.
Welcome to the year of the snake.
Jemimah Steinfeld
CEO, Index on Censorship
