I asked DeepSeek about China – then watched it censor itself midway through the answers

The arrival of DeepSeek, a new Chinese chatbot to rival OpenAI, Google and Meta, has sent shockwaves through the AI world and the US stock market.

The chatbot, which is purportedly more efficient and cheaper to run than its rivals, sent the stocks of chip-maker Nvidia crashing this week, and $938 billion was wiped from its value in a single day.

Road tests of DeepSeek were quick to prompt censorship concerns. There was a refusal to answer questions about controversial topics in China such as the Tiananmen Square massacre, which indeed I experienced when I used it for the first time.

Watch the video below to see DeepSeek’s real-time self-censorship in action.

I then asked it some other questions I didn’t expect DeepSeek to answer at all. What I noticed was strange. It did answer – before promptly deleting its own responses.

In this video, we take a look at what DeepSeek has to say about China’s history of human rights abuses, pro-democracy protests in Hong Kong — even a glowing review of Martin Scorsese’s “visually stunning” biopic of the Dalai Lama, Kundun (banned in China).

That is, until the responses suddenly disappeared.