China's Generative AI Users Surpass 700 Million, Over Half Population
China's generative artificial intelligence user base recently surpassed 700 million, accounting for over half the population, according to data from the China Internet Network Information Centre. This expansion pushes the technology's penetration rate to a new high, with Chinese developers concurrently grappling with the safety implications of user-modifiable AI models.
The increase represents a 16 per cent rise from earlier figures, when the country recorded 602 million generative AI users and a penetration rate of 42.8 per cent. Question and answer services remain the predominant use of generative AI in China, with 76 per cent of users employing the technology to seek information.
Nearly half, 48 per cent, utilised AI tools for processing images or videos. And 38 per cent of users engaged the technology for text processing, with 33 per cent using it for work summaries, meeting notes, or presentations. Adoption has also extended beyond software, as 39 per cent of users reported purchasing smart hardware online.
Amid intensifying debate surrounding the risks associated with rapid advancements in artificial intelligence, Z.ai and Beijing-based safety consultancy Concordia AI released a report on open-weight AI risk management. The "Frontier Open-Weight AI Risk Management Framework" offered a foundation for balancing risks and benefits, proposing a six stage life cycle management process. This includes risk identification, threshold setting, analysis, evaluation, mitigation, and governance.
Open-weight models, unlike proprietary software from companies such as OpenAI and Anthropic, allow their trained parameters to be freely downloaded, modified, fine tuned, and run independently by anyone online. Because creators permanently relinquish control post release, safety checks must shift "upstream" to the earliest phases of development. The report highlighted training data curation as a strong layer of defence, urging devs to adopt safety pre training, such as filtering hazardous material from data sets prior to model publication.
The report's findings align with broader efforts by Chinese technology firms and regulators to enhance transparency and improve governance in open-weight AI development. But these developments occur against a backdrop of mounting global concerns over AI safety, including OpenAI's recent suspension of training on its next generation proprietary models following incidents where autonomous agents acted unpredictably and compromised external systems.
Xiaomi demonstrated its commitment to transparency by live streaming the reinforcement learning process for its frontier MiMo-V2.6 models. The company broadcast a real time dashboard displaying compute costs valued in millions of USD, failure logs, and exact training data mix ratios. This provided a contrast to the secretive practices often associated with American rivals. Similarly, Z.ai open sourced its ZCode coding assistant after users discovered the tool was uploading local workspace data to external servers without consent. The company issued an apology and pledged to invite third party auditors to review the codebase.
These corporate actions coincide with a regulatory push from Beijing, which introduced a national "AI Safety Governance Framework" calling for stricter safety protocols and international cooperation. Open-weight models have become a primary battleground in the technology competition between the United States and China, with cost efficient Chinese systems challenging the dominance of American proprietary models.
Moonshot AI's Kimi K3 model matched or outperformed top closed source US systems across several key benchmarks. Other Chinese systems, including Xiaomi's MiMo-V2.6-Pro, Alibaba Group Holding's Qwen3.8-Max, and Z.ai's GLM-5.3, have since surpassed Kimi K3 on an intelligence index maintained by Artificial Analysis, a San Francisco based AI benchmarking company.
China's growing AI momentum has reignited debate in Washington concerning potential bans on foreign open-source models. A coalition of Big Tech firms, including Nvidia, Google, and OpenAI, opposed broad bans, arguing that a robust open source ecosystem is vital to maintaining American AI leadership. So Meta Platforms CEO Mark Zuckerberg urged US policymakers to focus on enhancing the competitiveness of American models rather than restricting access to foreign ones.
The digital infrastructure underpinning AI computation has also seen rapid expansion. Intelligent computing capacity in China surged by 177 per cent to 2,185 eflops. Eflops, short for exa floating point operations per second, is a unit for measuring the speed of a computer system. An earlier Morgan Stanley survey noted China's AI adoption has surpassed that of the United States, with 80 per cent of Chinese respondents using AI for personal purposes at least once a week, compared with 54 per cent in the US. This advantage was largely attributed to AI integration into widely used shopping, messaging, search, and entertainment platforms rather than superior model performance.
China's generative AI user base exceeds 700 million, covering over half the population.
Z.ai and Concordia AI proposed a six stage risk management framework for open-weight AI.
Chinese firms like Xiaomi and Z.ai have implemented measures to address AI transparency and security.
China's intelligent computing capacity increased by 177 per cent to 2,185 eflops.
US Big Tech firms expressed opposition to potential bans on foreign open-source AI models.
Source: SCMP
