China’s AI Race Is Moving Faster Than Its Safety Disclosures

The core finding from SemiAnalysis points to a significant gap between the speed of AI model development and the public communication of safety measures in China. With 857 models released by leading developers over several years, the 31 models with public safety disclosures represent a small fraction of the total. This suggests that while innovation is moving quickly, the transparency around how these models are being evaluated for potential risks is not keeping up.

This disparity could stem from various factors, including competitive pressures to launch models quickly, differing regulatory expectations compared to other regions, or a focus on internal testing without a strong incentive for public disclosure. For users and policymakers, the absence of public safety information makes it difficult to assess the risks associated with deploying these models, particularly as AI integrates into more critical applications.

The distinction here is between the pace of technological output and the pace of public accountability. While the volume of AI models indicates a vibrant and active development scene, the limited disclosures mean that external stakeholders lack a clear picture of the safeguards being implemented. This does not necessarily mean safety testing is absent, but rather that the results and methodologies are not being communicated broadly.

For companies operating within or alongside China's AI ecosystem, this trend highlights a potential area of divergence in industry practice. As AI models become more powerful and pervasive, the demand for clear, verifiable safety information is likely to grow globally. The current situation suggests that Chinese developers may need to consider evolving their disclosure practices to align with increasing international expectations around AI safety and transparency.