Huawei Co-Develops DeepSeek Model with Advanced Censoring Capabilities
Chinese technology giant **Huawei** has deepened its collaboration with **DeepSeek**, the country's fast-rising artificial intelligence startup, to co-develop the latest version of the
DeepSeek large language model. This project significantly increases the sophistication of automated Chinese internet censorship, distinguishing the model from its Western competitors[2][5].
The DeepSeek Model: A Homegrown AI Challenger
DeepSeek, backed by the Hangzhou-based hedge fund High-Flyer, entered the AI spotlight with its R1 model in early 2025. The model quickly drew attention for its ability to rival powerful Western AIs like
ChatGPT and
Gemini, reportedly trained at a fraction of their cost[3][5].
- **Strong regulatory alignment:** DeepSeek’s models are tightly aligned with Chinese government regulations on information control, especially political content[1][5].
- **Integration across industries:** The model has seen rapid adoption by major Chinese firms, including brokers, banks, and super-apps, indicating a broad impact on China’s digital economy[4].
Huawei: Supplying Hardware and Technical Expertise
Huawei provides DeepSeek with both hardware and engineering support, including its **Ascend**-series AI chips. These chips are used primarily for inference—a process where the AI responds to queries in real time—while model training has been challenged by persistent technical issues[2].
- **U.S. technology restrictions:** Following bans on the export of advanced
Nvidia H20 GPUs to China, DeepSeek was “encouraged” to use Huawei’s chips instead, which slowed progress due to reliability concerns[2].
- **New censorship technologies:** Huawei’s engineers have worked jointly with DeepSeek to strengthen content filtering and automatic moderation within the AI, effectively “improving” its ability to detect and censor politically sensitive subjects[1][5].
Global Reactions: Security Risks and Expanding Bans
The partnership has stoked security fears, especially among U.S. lawmakers and international privacy advocates, who warn that DeepSeek’s AI models could facilitate state surveillance and increase exposure to Chinese government influence[1][3][4].
- Government bans: The U.S. Federal government and several states—including Texas and New York—have banned DeepSeek on official devices, citing risks to data privacy and national security[3][4].
- Legislative action: Bipartisan legislation (“No DeepSeek on Government Devices Act”) aims to prohibit government use of DeepSeek, echoing earlier moves against Huawei equipment[3][4].
- Privacy complaints: Advocacy groups have filed complaints in both the U.S. and Europe regarding DeepSeek’s data handling and the risk of user information being shared with Chinese authorities[4].
Future Implications
As China seeks to
reduce its dependency on foreign AI technology, the Huawei–DeepSeek partnership may set a precedent for more tightly integrated, state-aligned AI systems. However, doubts remain about the performance and openness of these models when compared to Western counterparts[2][5].
For observers and AI users outside China, these developments signal a new phase in the **global competition** over advanced AI—and a sharp divergence in how different societies balance innovation, information control, and individual privacy.
