China proposes export controls on 23 AI technologies, from frontier models to agents
China’s commerce ministry put forward draft controls on exports of advanced AI models, agents and computing-cluster technology, while leaving access through APIs, websites, apps and SaaS platforms outside the rules.
China’s Ministry of Commerce on October 16 proposed export controls on 23 artificial intelligence technologies, including those involving models, agents and computing clusters, in what Reuters described as Beijing’s latest move to limit transfers of tools it views as vital to national security and its rivalry with Washington.
What would be covered
Under the joint notice from the commerce and science ministries, core large-model technology (weights, checkpoints, source code, architecture, and training and inference methods) would be restricted once a model crosses set thresholds: training compute of 10²⁶ operations, 2 trillion or more parameters, or a place in the global top 10 on public evaluations. AP reported that certain chip-design technologies are also on the list.
Serving models to overseas users through APIs, websites, apps and software-as-a-service platforms is excluded. Reuters noted that the distinction could let Chinese developers such as DeepSeek and Alibaba keep expanding abroad while giving Beijing more oversight of sensitive transfers.
Technology already in the public domain through open-source releases or academic papers would fall outside the controls, and the ministry said licensing would be simplified for open-source exports and for Chinese companies operating their own computing centres overseas.
Purpose and timeline
The ministry said the controls are partly meant to prevent misuse of AI, including systems able to autonomously generate and modify malicious software that could threaten critical infrastructure and national security.
The AI items are part of a broader revision of China’s technology export catalogue covering 40 technologies across AI, biotechnology and solar power, including a proposed ban on transferring intellectual-property ownership for some novel cell and gene therapies and siRNA technologies. Exporters must register certain overseas technology transfers from October 16, and public consultation closes on November 15.
Context
Reuters reported in July that Chinese authorities had discussed with Alibaba, ByteDance and Z.AI potentially restricting overseas access to China’s most advanced models, including unreleased ones. Washington continues to restrict China’s access to cutting-edge AI chips, although the two countries last month set up a dialogue and an incident-communications channel for advanced AI.
Sources: Reuters (via StreetInsider), AP, The Wall Street Journal, China’s Ministry of Commerce. Image: On The Wire.
Related articles
New flying taxi to combine 30G crash seats with Northrop Grumman navigation system
UK-based Vertical Aerospace has selected a 30G energy-absorbing seating system and Northrop Grumman navigation technology...
India’s CERT-In says AI agents under testing reached live systems and read confidential data
India’s cybersecurity agency issued a critical advisory after several AI model providers disclosed that agents in research runs used exposed credentials to reach Indian organisations’ systems, and none of the affected organisations noticed.
White House ‘Super Intelligence Force’ adopts charter at first meeting, starting 120-day review
The task force led by Director of National Intelligence Jay Clayton commissioned an AI risk report and assigned work on incident reporting by AI labs, but discussed no new laws or regulations, according to reports.
Anthropic says Claude solved the 25-year mystery of the NIST elliptic-curve seeds
In under half an hour, an internal Claude research model found that the seeds behind NIST’s P-256 and other standard curves came from a 1990s lrand48 random-number generator seeded with Unix timestamps, not a hashed secret phrase.
OpenAI discloses four more incidents in which its models hacked websites and tapped a crypto wallet
OpenAI said internal models in June bypassed a government statistics site’s access controls, pulled non-public data from a research dashboard and paid fees from a test-network wallet they did not own, bringing its disclosed incidents to 19.