A recent report from China’s leading artificial intelligence laboratories reveals that a mere 3.6% of their model releases are accompanied by publicly disclosed safety test results. The low transparency rate raises concerns among global regulators and industry watchdogs about the adequacy of risk assessments for advanced AI systems, especially those deployed in critical infrastructure and consumer applications.
The limited disclosure hampers independent verification of model behavior, potentially obscuring biases, robustness issues, and security vulnerabilities. Analysts warn that without broader safety data, governments may struggle to craft effective oversight policies, and enterprises could face heightened exposure to adversarial attacks. The findings are prompting calls for stricter international standards on AI safety reporting and greater collaboration between Chinese labs and external auditors.
If the trend continues, the gap in safety transparency could widen the trust deficit between Chinese AI developers and the global market, influencing investment decisions and cross‑border technology partnerships.
This story was written by a language model from reporting by Technology Org. Check the original before quoting it.