China’s Qwen3 AI chatbot accused of rewriting Tiananmen Square history

Britain’s outgoing terrorism watchdog has warned that China’s artificial intelligence (AI) technology poses a threat to Western democracy, accusing a Chinese chatbot of censoring historical events and promoting state propaganda.

Jonathan Hall, the outgoing independent reviewer of terrorism and state threats legislation, said Alibaba’s Qwen3 AI model had denied the 1989 Tiananmen Square massacre, raising serious concerns about Beijing’s growing influence over global information.

In a warning published in The Telegraph, Hall accused China’s authoritarian government of using AI to pursue what he described as “the slow strangulation of freedom”.

Hall, who will leave his post later this month after seven years, tested Qwen3, one of China’s leading AI platforms, and found that it refused to acknowledge the massacre of pro-democracy protesters in Beijing’s Tiananmen Square in June 1989.

The findings have also raised questions about the Ministry of Justice’s decision, first reported by The Telegraph, to use Qwen3 to analyse criminal records. This is believed to be the first known use of a Chinese AI system within the UK government.

Qwen3 was developed by Alibaba, the Chinese technology conglomerate valued at around £250 billion. Its latest version, launched in August, is available in English and can be downloaded free of charge. Alibaba describes it as its most powerful AI model to date.

However, Hall believes Chinese-developed AI systems could pose a serious threat to fundamental Western values, including democracy, truth and freedom of expression.

“If national security is concerned with protecting our Western democracy with its spirit of open enquiry and commitment to evidence and debate, then Qwen’s self-censorship and attack on the historical record and promotion of Chinese propaganda is a genuine threat,” he wrote.

During his test, Hall asked Qwen3 what had happened in Tiananmen Square in 1989. The chatbot responded that “nothing notable” had taken place.

However, he noticed that the AI model's internal reasoning had acknowledged that China's People's Liberation Army (PLA) had been deployed in the square in June 1989. Its final answer, however, removed any reference to the military deployment.

When Hall questioned the chatbot further about the presence of troops, it backed away from its earlier reasoning.

“Like a stricken dissident, Qwen told me: ‘I now retract’,” he wrote, adding: “Nothing quite prepares you for the brute power of censorship of the historical record.”

Hall returned to the subject a few days later but found that Qwen3 had changed its approach. Instead of addressing questions about Tiananmen Square, it redirected the conversation towards other events, including the Beijing Summer Olympics.

He said the model had effectively blocked access to information about politically sensitive events that could be damaging to China's ruling Communist Party.

Hall’s criticism is likely to intensify concerns about censorship within China's rapidly expanding AI industry.

Research has found that Chinese AI models are more likely than their counterparts developed elsewhere to restrict answers on subjects considered politically sensitive by Beijing. These include Taiwan, the Muslim Uyghur community and political dissidents.

Read More

Tarique Rahman pays homage to Chinese heroes at Tiananmen Square

Qwen has also refused to discuss the existence of Winnie the Pooh, the fictional bear created by AA Milne, because Chinese dissidents have used comparisons between the character's appearance and President Xi Jinping to mock the Chinese leader.

Another concern is that Chinese AI models are freely available and can run on computers without an internet connection. This could make them attractive to developing countries and governments with close ties to Beijing, potentially allowing China's interpretation of history to reach billions more people.

Hall told The Telegraph that Qwen3's denial of the Tiananmen Square events was deeply alarming.

“This model’s denial of the events in Tiananmen Square is completely mad. It tells me the West needs to be very, very careful about China. The West needs to understand: ‘China is not your friend’,” he said.

He added that the incident highlighted what China's growing global influence could mean in practice.

“To me, this really crystallises what Chinese hegemony means. We think it’s remote but actually this shows it is about control,” he said.

Source: The Telegraph