What happened
On 16 October 2026 China's Ministry of Commerce (MOFCOM) and Ministry of Science and Technology (MOST) issued Announcement [2026] No. 48, seeking public comment until 15 November 2026 on proposed amendments to the Catalogue of Technologies Prohibited or Restricted from Export, adding 40 technologies across AI, biopharma and photovoltaics. The AI additions span 23 technologies: AI large model technology, AI agent technology, AI computing cluster technology, high-performance detection and information countermeasures (cyber offense/defense), and information processing (speech/voiceprint recognition, Chinese-language translation). Prohibited vs restricted categories turn on control parameters (e.g. core vs important data used for training, training compute >= 10^26 FLOPs, >= 2 trillion params, top-10 benchmark capability, 10,000+ AI accelerator cluster). Open-source/academic-public-domain releases and API/SaaS/on-device access to domestic models are carved out. A temporary pre-export registration requirement took effect immediately from 16 October 2026.
Why it matters
This materially changes the regulatory landscape for AI technology transfer and security for any multinational with AI R&D or collaboration in China, and directly controls AI-enabled cybersecurity and cyber-offence capabilities. It establishes precedent for risk-based, parameter-threshold technology controls and an automatic-licensing mechanism, and applies immediately via registration even before the catalogue is finalised.
Action needed
Multinationals with AI R&D, model/agent or compute-cluster technology in China should assess export-classification exposure (training data classification, compute thresholds, agent/cyber capabilities), set up MOFCOM pre-export registration for affected exports, and consider submitting comments before 15 November 2026.