By Hamza Faisal
For years, the chip war ran in one direction: Washington cutting China off from Nvidia's most advanced GPUs while Beijing raced to build domestic alternatives. On Friday, China flipped the script — drafting a plan to restrict exports of its own AI technology, from trained models and intelligent agents to chip designs and computing clusters.
The Ministry of Commerce and the Ministry of Science and Technology published a draft revision — Announcement No. 48 — to the catalogue of technologies prohibited or restricted from export. It proposes adding 40 technologies across AI, biomedicine and photovoltaics: 14 for an outright export ban and 26 for restricted export under license.
Public comment runs until November 15.
Of the 40, 23 are AI-related: large language models, intelligent agents, chip design and computing clusters. AI technology that could generate and mutate malicious code is on the proposed outright-ban list, alongside perovskite solar-cell manufacturing tech and some cell and gene therapy platforms.
The paperwork starts now. China imposed an interim pre-export registration requirement effective immediately, lasting until the revised catalogue takes effect.
It covers technology-ownership transfers, overseas licensing, and ownership changes that place control of China-based technology with foreign entities — including transactions already underway but not yet completed.
There are carve-outs: technologies in the public domain through published papers or open-source releases are excluded. A ministry spokesperson said most AI technologies would remain freely exportable and described the scope as "precise and limited."
A wall facing the other direction
The move is a milestone: the first time Beijing has moved to restrict software and trained AI systems themselves, rather than just the hardware underneath them.
The Wall Street Journal reported in July that Chinese officials had held discussions with domestic AI labs about protecting proprietary technology, worried that sharing it could help adversaries and ultimately be weaponized against China.
The target, according to the Financial Times, is Chinese open-weight AI — models anyone can download and run. Alibaba's Qwen, ByteDance's Doubao, Zhipu's GLM and DeepSeek's releases have become the cheap default for developers avoiding OpenAI and Anthropic prices.
Chinese open-weight models captured 17.1% of global open-weight downloads in the year to August 2025, overtaking the United States' 15.8% share for the first time.
"Routine," Beijing says — the timing says otherwise
The ministry insists the revision is routine and targets no specific country, region or entity, aiming to keep AI technology "safe and controllable."
On Chinese AI companies expanding abroad, the spokesperson said it would hinge on "whether it involves the transfer of restricted technologies from within China to locations outside its borders."
Beijing has also been moving on people: in September, exit bans for tech engineers took effect under Decree 841, letting authorities stop chip, rare-earth and battery engineers from leaving.
Friday's draft goes after the technology instead — with solar and biomedicine riding along: 12 solar items covering polysilicon, wafer and cell manufacturing, and five biomedical items on cell and gene therapy.
What happens next
Public comments close on November 15, and the interim registration regime means every qualifying transfer — including deals already mid-flight — must be logged with regulators in the meantime. The ministry has not said how many deals the interim rule has already caught.
For companies that quietly built products on top of Chinese open-weight models, the draft is a warning shot: the cheap foundation could get a regulator's veto. For Beijing, the logic now mirrors Washington's: the technology at the center of the century's most consequential race is too strategic to move across borders unchecked.
The chip war has a second front — and this one is made of software.
By Hamza Faisal
