A comprehensive analysis of artificial intelligence deployment patterns in China reveals a stark disparity between the volume of model releases and the transparency of safety practices. According to AI Weekly, researchers examining output from nine major Chinese AI developers found that the overwhelming majority of new systems enter the market without any disclosed safety assessment.
The Numbers Tell a Troubling Story
Between 2021 and September 2026, the tracked companies collectively released 857 AI models. Of these, only 31 included a published safety evaluation. Nine of those evaluations were available at the time of launch. The remaining 813 releases, representing 94.9 percent of all deployments, shipped with no safety disclosure whatsoever.
The cohort examined includes both established hyperscalers and innovative startups: ByteDance, Alibaba, Tencent, and Baidu represent the large incumbent players, while DeepSeek, MoonShot, Zhipu, MiniMax, and StepFun represent the emerging competitive tier. Each of these organizations contributed to a pattern that suggests safety documentation is not yet a standard part of the release workflow in the region's AI industry.
What This Means for the Industry

The research underscores a fundamental disconnect between development pace and risk management practice. While Chinese companies have positioned themselves as serious competitors in frontier AI capabilities, the apparent absence of systematic safety evaluation and public reporting diverges sharply from the approach taken by American counterparts like OpenAI, Anthropic, and Google.
This gap has implications beyond corporate responsibility. Regulators worldwide have begun scrutinizing how companies approach AI safety before deployment. The absence of third-party or published internal assessments complicates efforts to benchmark risk across geographies and competitive contexts. It also raises questions about whether market pressures, regulatory frameworks, or internal priorities shape these disclosure patterns.
Context Matters
The findings should be understood within the broader context of AI governance in China and the dynamics of a rapidly moving competitive landscape:
- China's regulatory approach to AI emphasizes content control and national security more heavily than safety testing infrastructure
- Speed to market may take priority over documentation in hypercompetitive segments
- International transparency standards for AI safety remain nascent and unevenly adopted
- Published safety evaluations may not reflect actual internal testing practices
The research does not necessarily imply that Chinese developers are neglecting safety work internally. Rather, it documents a transparency gap: whether or how much safety evaluation occurs behind closed doors remains opaque to external observers and regulators alike.
Implications for Global AI Governance
This analysis arrives at a moment when policymakers in the United States, European Union, and other jurisdictions are debating how to mandate or incentivize safety practices. The Chinese case suggests that competitive pressures and regulatory environments shape not just the substance of safety work but also decisions about whether and how to document it publicly.
The data point to a potential fragmentation in global AI governance: regions or companies that prioritize disclosure may face efficiency trade-offs relative to competitors that do not. Resolving this dynamic will likely require coordination among major markets and clearer international norms around what constitutes responsible AI release practices.



