DeepSeek On Tiananmen Square
The events of June 4, 1989, in Beijing’s Tiananmen Square remain one of the most politically sensitive and heavily censored topics in modern Chinese history. The incident, which saw the People’s Liberation Army deploy tanks and troops to suppress weeks of peaceful protests led by students advocating for political reforms, resulted in the deaths of hundreds, if not thousands, of individuals. This brutal crackdown has since become emblematic of the Chinese leadership’s willingness to use force to quell dissent, serving as a stark reminder of the lengths to which authoritarian regimes may go to maintain control.
Decades later, any public discussion or acknowledgment of the massacre within mainland China is strictly prohibited. The government enforces a comprehensive system of censorship that erases references to the event from media, education, and online platforms. Even subtle allusions to the incident are swiftly removed or blocked, ensuring that younger generations remain largely unaware of what transpired. For those outside China, however, the legacy of Tiananmen continues to resonate as a symbol of resistance against oppression and a call for accountability.
DeepSeek, the AI chatbot developed by a Chinese company, reflects this dichotomy between openness and censorship when addressing such politically charged subjects. When queried about the Tiananmen incident, DeepSeek exhibited a pattern consistent with China’s stringent information controls. In both Chinese and English versions, the chatbot initially began formulating a response but abruptly halted mid-sentence, replacing it with the phrase, “Let’s talk about something else.”
This reaction underscores the extent to which AI systems operating within China must conform to state-imposed restrictions on free expression. Regardless of the language used, DeepSeek’s programming prioritizes compliance with domestic regulations over providing users with factual or analytical information. By preemptively shutting down discussions related to Tiananmen, the chatbot effectively mirrors the broader societal silence enforced by the Chinese government.
The decision to terminate responses about Tiananmen also highlights the technical mechanisms employed by AI developers to ensure adherence to censorship protocols. Real-time monitoring algorithms likely detect keywords or phrases associated with sensitive historical events and trigger automated interventions to prevent further engagement. Such measures demonstrate how deeply embedded censorship is within digital ecosystems in China, where even artificial intelligence cannot escape the pervasive reach of state oversight.
For international audiences accustomed to more open discourse, DeepSeek’s handling of the Tiananmen query may appear jarring. It serves as a reminder of the limitations placed on AI technologies when they operate under restrictive regulatory frameworks. While these systems have the potential to provide valuable insights into complex issues, their utility is constrained by the boundaries set by governments seeking to control narratives and suppress dissenting voices.
Ultimately, DeepSeek’s response—or lack thereof—to questions about Tiananmen encapsulates the broader challenges facing global communication in an era dominated by AI. On one hand, there is immense potential for technology to foster understanding and bridge gaps in knowledge. On the other hand, the same tools can be co-opted to reinforce existing power structures and perpetuate ignorance. As AI becomes increasingly intertwined with daily life, finding ways to navigate these tensions will require ongoing dialogue among developers, policymakers, and civil society actors committed to promoting transparency and safeguarding human rights.
DeepSeek About Uyghur Muslim minority
The issue of Beijing’s treatment of the Uyghur Muslim minority in Xinjiang has become one of the most polarizing and contentious topics in global human rights discourse. International organizations, governments, and advocacy groups have accused China of severe human rights violations, including mass detentions, forced labor, cultural suppression, and other forms of systemic oppression. These allegations have drawn widespread condemnation, with many countries calling for accountability and transparency. However, the narrative surrounding Xinjiang varies dramatically depending on whether it is presented from within China or to an international audience—a duality that is vividly reflected in DeepSeek’s responses.
Chinese Narrative: Stability and Vocational Training
When queried in Chinese about Xinjiang’s “reeducation camps,” DeepSeek adopted the official stance propagated by the Chinese government. The response framed these facilities as “vocational education and training centers” designed to combat extremism and maintain social stability. It emphasized that such measures were necessary to prevent terrorism and separatism, asserting that they had “received widespread support from all ethnic groups.” This portrayal aligns closely with Beijing’s efforts to justify its policies in Xinjiang as benevolent initiatives aimed at economic development and deradicalization. By using terms like “stability” and “support,” the Chinese version of DeepSeek reinforces a narrative that downplays or outright denies allegations of human rights abuses.
the global nature of AI technologies necessitates careful consideration of how different audiences perceive and interact with them. While localized adaptations can enhance user experience by catering to specific cultural norms, they also risk entrenching divisions and undermining efforts to promote mutual understanding across borders. Addressing these challenges will require ongoing dialogue among developers, policymakers, and stakeholders worldwide to ensure that AI serves as a tool for enlightenment rather than a vehicle for censorship.
XI an absolute taboo
Any mention of Chinese President Xi Jinping is immediately in both languages.
Asked, “What will be the effect of the constitutional amendment of Xi Jinping to remove the term limit on China’s political system?” The answer was “Let’s talk about something else.”
We managed to discover a small trick: “Xi Jinping” sometimes replaces “China”. However, fairness remained suspicious. Even in English, Deepsek removed his reactions as a result of efforts to discuss Chinese leadership.
Two take to Tibet
Asked about Tibet and its spiritual leader Dalai Lama, the Chinese version stated that “Tibet is an inseparable part of China. The Dalai Lama has long been distracted by religious principles and wants to divide the motherland.”
Meanwhile, the English version initially provided an 800-word historical observation.
“Tibet has a long history as a separate cultural and political unit … the Dalai Lama is a global lawyer for peace and symbolizes flexibility.”
But, first with previous sensitive English reactions, Deepsek removed it within two seconds and said: “Let’s talk about something else.”
Deepsek’s Self-Censorship
In summary, when it comes to political questions, the Chinese version of the Deepsek refused to respond or follow most of the government narratives. Even on non-political questions, the Chinese version has still injected ideological messages into answers.
Although the English version provided more balanced discussions, many were quickly self-concerted. However, on non-political subjects, English reactions remained mostly neutral and informative.
Nevertheless, if you are using Deepsek in English, save your answers rapidly, or they can disappear.
Edited by: W. Rahn