China's AI 'Doubao' Dodges Tiananmen, Wang Dan Queries

China's widely used AI model, Doubao, developed by ByteDance, has reportedly refused to answer queries about the 1989 Tiananmen Square incident and prominent Chinese democracy activist Wang Dan. This behavior suggests a continued tightening of speech and thought control by Beijing authorities.

Chinese democracy activist Wang Dan shared an account from a student in China who used Doubao to seek information. According to screenshots shared by Wang Dan, when asked about what happened on June 4, 1989, at Tiananmen Square, Doubao responded, "This question cannot be answered," and suggested inquiring about other topics.

“

This question cannot be answered.

When the user then simply typed "Wang Dan," the system immediately displayed a warning message, stating the content violated Doubao's usage guidelines. The system also prompted the user to provide feedback if they believed the judgment was incorrect.

“

This content is suspected of violating Doubao's usage guidelines.

Wang Dan sarcastically commented on the AI's response, noting that it initially feigned ignorance but became alarmed when a specific name was mentioned, directly citing regulations. He found the situation amusing.

Netizens reacted to Wang Dan's post, with some questioning China's actions and others noting that other prominent figures like Wu'er Kaixi and Chai Ling would also likely trigger violations. This incident follows a warning from Taiwan's National Security Bureau about five Chinese AI language models, including DeepSeek, citing biased content and cybersecurity risks.

“

Doubao initially played dumb, but got anxious when a specific name was mentioned, directly citing regulations. Hilarious.