This is the new MLex platform. Existing customers should continue to access MLex here until migrated to the new platform. For any migration queries, please contact Customer Services or your Account Manager.

China treads carefully on AI legislation amid US rivalry, industry stress on growth

By MLex Staff

(November 6, 2024, 19:37 GMT | Comment) -- As the artificial intelligence race intensifies globally, China is taking a measured approach to legislation, mindful of the ongoing tech rivalry with the US and industry demands for development flexibility.

China's top legislature has embraced the mantra of "tolerant, cautious and gradual" for AI regulation, highlighting a need to further build consensus. This approach was aired in a recent speech by Wang Hongyu, director of the Research Office of the Legislative Affairs Commission of the National People's Congress Standing Committee, the office orchestrating research for a potential AI law.

This strategy aims to strike a balance between governance, industrial growth and global competitiveness. It also echoes the regulatory philosophy of "tolerance and prudence" that once propelled China's Internet boom. Officials are now applying similar principles to AI, promoting the idea that "lack of development is the greatest insecurity."

Industry leaders have pointed to US restrictions on advanced AI chips and China's tech gap with US rivals in arguing that development should take precedence over regulation. They contend that existing laws suffice for oversight, citing the Cybersecurity Law, Data Security Law, Personal Information Protection Law, along with the suite of rules on recommendation algorithms, deep synthesis and generative AI.

Lawmakers appear receptive to these industry arguments. In the speech, Wang suggested using existing legal frameworks or judicial interpretations to address AI-related legal challenges, including the fair use of intellectual property rights in the training of large language models, or LLMs (see here). He advocated a "small, flexible, and fast" legislative approach to swiftly address pain points, contrasting with the EU's comprehensive AI law.

China's pro-growth strategy is evident in changes to its draft interim rules for generative AI services. The initial proposal included challenging obligations for AI providers that could have overburdened market players. Subsequent modifications softened or removed many of these duties (see here and here).

— Existing and potential frameworks —

China's AI governance spans multiple agencies, including Internet, industry, tech, security and science regulators. These departments often pursue divergent agendas.

A major focus of the current framework is content regulation, primarily overseen by the Cyberspace Administration. It prioritizes information content safety, followed by consumer protection and market competition order.

For instance, China's interim rules on generative AI services strictly limit content creation. Top prohibitions include content that could subvert state power, overthrow socialism, or endanger national security. Secondary requirements address preventing discrimination, safeguarding competition and protecting data subjects' rights.

As for consumer-facing LLMs, authorities have already employed a record-filing system — effectively a vetting process — to regulate providers and technical supporters. This system requires filing algorithm information with the national Internet regulator and model details with local regulators. Before approving AI products or services, regulators seek to understand service types, goals, scenarios, R&D, computational resources, training data and security assessments.

As China develops broader AI legislation, two government-linked academic groups have proposed frameworks. Generally, Chinese regulators are understood to be reluctant to fully emulate the EU’s risk-based AI governance approach.

One proposal envisages an AI management system that combines a negative list and a record-keeping system to address potential risks (see here). The other proposal suggests a tiered regulatory framework for AI and special-application scenarios, with additional obligations proposed for "critical AI" developers and providers (see here).

While China has yet to take an official stance, the Internet regulator recently endorsed an AI safety governance framework outlining a tiered, category-based management system, offering an insight into its regulatory thinking (see here).

— Whole-of-society efforts —

China prioritizes specialized industrial AI applications over broader uses such as LLMs, targeting sectors such as industry, agriculture, education and healthcare. This strategy reflects the government's view of AI as a "new productive force" driving innovation, productivity and industrial competitiveness.

In March, the government’s work report launched an AI-plus campaign aimed at advancing the industrial use of AI technologies. The Ministry of Industry and Information Technology, or MIIT, is spearheading the initiative, which will include the automotive sector (see here).

To support these efforts, the National Data Bureau is collaborating with MIIT to create a national unified computing system, ensuring adequate access to computing power. Additionally, the bureau is developing foundational data systems to facilitate the supply, flow and trade of data.

In another display of a “whole-of-society" commitment, massive investments are being channeled into the AI sector. Major cities such as Shanghai, Beijing, and Shenzhen — all vying for the title of "AI City" — have established AI industry parent funds or government-led funds to boost development.

Government venture capital funds often take the lead in investing in AI companies, setting precedents that private ones tend to follow. Other supportive mechanisms include subsidies, such as those provided in Shenzhen (see here).

To date, five Chinese generative AI startups, Zhipu AI, Moonshot AI, MiniMax, 01.ai and Baichuan AI, lead China’s race to rival US counterparts such as OpenAI and Anthropic. But no clear winner has emerged yet in foundation models. The startups also compete with domestic Internet titans, such as Alibaba, Baidu and TikTok’s owner ByteDance.

— Realigning national goals —

China's AI push comes at a cost. The swift AI integration across sectors has raised concerns about job losses and public discontent, notably in autonomous driving.

Recently, Baidu Apollo's rollout of driverless robotaxis in the Chinese city of Wuhan has drawn opposition from local taxi operators (see here). Their calls for stricter controls on Apollo Go sparked debate about balancing China's tech ambitions with job preservation.

The local government, however, downplayed the issue, and Beijing remained silent on the outcry at the time. This reaction underscores China's firm focus on achieving technological self-sufficiency and competing with the US, particularly in areas deemed crucial for future economic growth.

As joblessness mounts amid economic headwinds, however, the Chinese government is recalibrating its stance. A more nuanced approach to the adoption of AI technology is emerging, one that weighs social consequences alongside technological ambitions. The recent policy jointly issued by China’s Central Committee of the Communist Party and the State Council has called for “actively addressing” the impact of new technologies on jobs (see here).

The government has touted a "human-centered" approach to AI governance, a mantra that seemed tailored more for foreign audiences than domestic application. But as unemployment grows and social fissures widen, the government finds itself compelled to rewrite its playbook. The previous paean to productivity, innovation and competitiveness in AI development and regulation now finds itself in an uneasy tango with social considerations. 

— Analysis by Wang Juan and Yonnex Li.

Please email editors@mlex.com to contact the editorial staff regarding this story, or to submit the names of lawyers and advisers.

Tags