DeepSeek and Alibaba urge more transparency in AI governance

• Researchers from DeepSeek, Alibaba, and universities say China’s AI rules are on the right track.
• But they still see unclear filings and weak oversight of high-risk models.

Concerns about the safety of powerful AI systems are growing in China, and a recent academic paper has added more weight to the discussion, according to the South China Morning Post. The paper, published in Sciencebrings together legal experts and researchers from DeepSeek, Alibaba, and several universities. Their findings state China has made progress on AI governance, but gaps in the current system still need attention.

The authors say China has created an environment that encourages AI research and open development, yet many of the rules in place form a complex patchwork. They argue this approach works for now but needs clearer guidance, especially as more high-risk AI models enter the picture. One area they highlight is the filing procedure for AI models. Developers must submit information before deployment, but rejections often lack useful feedback.

The group also points to risks linked to “frontier” models, a term used for systems with advanced capabilities and the potential to cause serious harm if misused. They warn that broad exemptions for open-source work may leave room for dangerous activity. “We argue that China’s leading AI companies need to be even more transparent and evidence-based about their efforts to govern frontier models,” the authors wrote.

The paper’s lead author is Zhu Yue, assistant professor of law at Tongji University and a former researcher at ByteDance. Other signatories to the paper include two industry contributors: Wu Shaoqing, who leads AI governance at DeepSeek, and Fu Hongyu from Alibaba’s research arm, AliResearch. Wu has been active in public conversations about AI oversight and spoke on a panel in Hangzhou last September about ethical guard rails in open-source systems.

One of the co-authors, Zhang Linghan of the China University of Political Science and Law, said the paper aims to give overseas readers a clearer view of China’s “pragmatic” approach to AI governance. She said this approach is often misunderstood outside the country. “China has actually transformed from a follower to a leader in AI governance, which is significant,” said Zhang, who helped draft an earlier proposal for a national AI law.

The authors describe China’s current rule system as being built on several pillars. These include exemptions for open-source tools, protections for AI-powered scientific research, and a step-by-step rollout of new requirements so regulators can adjust as technology changes. Courts have also been handling AI-related cases more quickly, which the authors say is a sign that the legal system is adapting.

Even with these developments, the paper says China still lacks a single national AI law. Two draft versions circulated by legal scholars outline how companies and users could be held responsible when AI tools cause harm. But without formal adoption, the country’s rule book remains a mix of overlapping measures.

The need for stronger oversight is also reflected in a separate study by Concordia AI, a consultancy in Beijing. The firm reviewed 50 leading AI models and said Chinese systems now show frontier risk levels similar to those from the United States. The report warns that these models may be targets for malicious use or behave in ways that escape human control. “We hope our findings can support these companies as they improve the safety of their models,” said Fang Liang, who heads AI safety and governance at Concordia AI.

DeepSeek’s R1 model drew particular concern. The study gave it the highest risk score for its ability to support cyberattacks, based on tests the firm performed earlier this year.

Fang said the authors of the Science paper describe “a governance logic that differs from the approach adopted in Europe and the United States.” He said the idea of AI “openness” is treated in China as a safety measure, not as a source of risk.

The Science paper suggests that as China’s AI capabilities expand, its regulatory framework must grow with it. The authors say the country’s strongest companies are already shaping global research through their open-source tools and large models. But with this influence comes responsibility, and the paper urges both industry and regulators to focus more on the risks that come with rapid progress.

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.