Dmytro Kharkov

China drafts AI and chip export curbs to tighten control over technology transfers

China drafts AI and chip export curbs to tighten control over technology transfers
China tightens tech exports

Beijing is moving to widen its export control regime to cover advanced artificial intelligence technologies and chip design tools as competition with the U.S. intensifies. The draft measures target transfers of leading AI models trained on sensitive Chinese data and certain advanced semiconductor-related technologies, while carving out exemptions for some overseas operations and open-weight models.

Highlights

  • China's draft rules restrict exports of advanced AI models, agents, and chip design technologies, tightening oversight of overseas technology transfers in response to regulatory gaps.
  • The new controls target technology transfers through foreign acquisitions rather than products, with exemptions for open-weight AI models unless they exceed 2 trillion parameters or are top global performers.
  • Simplified approval procedures would enable Chinese firms like ByteDance, Alibaba, and Tencent to operate overseas data centers more flexibly as U.S. trade restrictions limit their access to advanced chips.

Draft controls broaden oversight of AI and chip technologies

As reported by Financial Times, citing the Chinese commerce ministry, the draft rules issued jointly with the ministry of information technology would restrict exports of leading AI models trained on sensitive Chinese data, as well as advanced AI agents used in areas such as critical digital infrastructure, defence and government services.

The proposed expansion of China’s lists of prohibited and restricted export technologies would give Beijing greater oversight of overseas transfers of advanced AI capabilities. The move is partly aimed at closing what Chinese authorities view as a regulatory loophole exposed by Meta’s $2 billion acquisition of AI app Manus, a deal that regulators later ordered to be unwound.

The draft controls also cover technologies used to design advanced processing and memory chips, along with packaging and networking equipment. The restrictions focus on transfers of underlying technology overseas, including through acquisitions by foreign entities, rather than automatically extending to products built using those technologies.

The rules include exemptions for open-weight AI models that are made publicly available and are not trained on sensitive data. Still, China would retain the ability to restrict transfers of technology tied to such models if they exceed 2 trillion parameters or rank among global top performers, potentially limiting offshore transfers of ownership or key employees.

Industry implications for Chinese tech groups

The parameter threshold highlights the scale now emerging in China’s AI sector. Moonshot’s K3 model, which has drawn global attention this year as its performance approaches top-tier offerings from Anthropic, has 2.8 trillion parameters.

Simplified approval procedures would apply to Chinese companies building and operating AI data centres overseas, manufacturing their own chip designs abroad, conducting research with foreign partners or helping set international standards. That could offer some operational flexibility to groups such as ByteDance, Alibaba and Tencent as they expand overseas data centre capacity while U.S. and allied restrictions limit access to advanced chips within China.

The commerce ministry says the proposed controls would ban access to AI models trained on sensitive core data, but would not affect indirect overseas access to other Chinese AI models through apps or websites. It says the key distinction is whether restricted technology itself is being transferred from China to another country.

The latest proposal also extends beyond AI and semiconductors to some biomedical and solar technologies. It follows China’s 2025 revision of its export restriction list, which added several lithium-ion battery manufacturing technologies and built on existing controls in sectors including rare earth extraction and processing.

In our earlier article on ByteDance’s AI safety push, we covered how its Seed research unit set up an Alignment & Preparedness team and launched a cross-border hiring drive to expand testing, monitoring and risk-readiness for increasingly autonomous AI systems. We noted that the move signaled AI safety and oversight becoming operational requirements as scrutiny of advanced models and agents rises in both China and the U.S.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.