Aleph Alpha study: Chinese models often avoid sensitive topics or repeat official positions
What happened
On 4 October 2026, The Decoder reported that German AI company Aleph Alpha developed a benchmark testing Alibaba's Qwen, DeepSeek and Moonshot AI's (Kimi) models with 967 sensitive topics. According to its own scoring system, only 17% to 41% of answers were rated balanced, while the rest repeated official positions, avoided the question or refused to answer. The report did not disclose the specific topics tested, details of the scoring criteria or the individual scores of each model.
Written by AI from the reports · updated 12 d ago
Timeline
Follow the reports to see the event from each side.
- The DecoderChinese AI models parrot state doctrine or refuse to answer on sensitive topics
Chinese AI models often follow the party line on politically sensitive questions, according to an Aleph Alpha study that rated only 17 to 41 percent of answers as balanced. Aleph Alpha sells "sovereign AI" to governments, giving it a commercial interest in distinguishing its models from Chinese competitors.