New Delhi, Aug 5 (IANS) Chinese AI models like DeepSeek have a serious limitation due to censorship and are distrusted by many governments and users worldwide, according to an article published by East Asia Forum. When DeepSeek’s 'R1' model arrived in January 2025, it outperformed Western competitors in Maths and coding but refused or deflected roughly 85 per cent of prompts concerning politically sensitive topics in China. The 2023 Interim Measures for the Management of Generative AI Services requires any public-facing model to uphold ‘core socialist values’ and pass a state security review. The article observes that censorship persists even when DeepSeek’s open-weight models are run privately, and research shows that foundation models originating from China have higher refusal rates, shorter responses and less accurate answers to a battery of 145 questions about Chinese politics than models built elsewhere. Chinese state control of the media seeps into training data, steering large language models towards more favourable portrayals of states with tightly controlled media. While developers often adopt Chinese open-source models for comparable performance at lower cost, most large enterprises contract with US providers or run inference through major US cloud platforms. DeepSeek’s consumer app has seen downloads tumble from an estimated 171 million in 2025 to an estimated 35 million in the first half of 2026. Australia, Taiwan and Italy have barred it from use in government systems.