China's MOFCOM and MOST propose adding 23 AI technologies, including frontier model weights and agents, to export control catalogue, with interim registration now in effect
Edited by Elliot Marsh, Editor · written with AI from the linked sources (how)
On 16 Oct 2026 China's Ministry of Commerce and Ministry of Science and Technology issued Announcement No. 48. It proposes adding 40 technologies (14 prohibited, 26 restricted; 23 in AI, 5 biomedicine, 12 photovoltaics) to the Catalogue of Technologies Prohibited or Restricted from Export. The draft is open for comment until 15 Nov 2026. AI items include model weights and related technology at 10^26 FLOPs, 2T parameters or global top-10 standing, plus agents, offensive-cyber AI, compute clusters and chip design. API/SaaS access is not counted as export. An interim pre-export registration requirement through MOFCOM's platform applies immediately to listed technology transfers, ownership/control changes and overseas licensing.
Entities: Ministry of Commerce of China (MOFCOM), Ministry of Science and Technology of China (MOST), Cloud Security Alliance AI Safety Initiative, ChinaTalk, Catalogue of Technologies Prohibited or Restricted from Export
0 primary
What happened
On 16 Oct 2026 China's Ministry of Commerce (MOFCOM) and Ministry of Science and Technology issued Announcement No. 48. It proposes adding 40 technologies to the Catalogue of Technologies Prohibited or Restricted from Export: 14 prohibited and 26 restricted, of which 23 are AI. The AI items reportedly cover model weights and related technology at 10^26 FLOPs, 2 trillion parameters or global top-10 standing, plus agents, offensive-cyber AI, compute clusters and chip design. The catalogue additions are a draft open for comment until 15 Nov 2026. A pre-export registration requirement on MOFCOM's platform is said to apply immediately to listed transfers, ownership or control changes and overseas licensing. API and SaaS access is reportedly not treated as export. The details come from a secondary analysis, not the announcement text.
Why it matters
If finalised, companies moving frontier model weights or related know-how out of China, or licensing them abroad, would need to register or seek approval. Multinationals with Chinese AI teams, local deployments or ownership changes are most exposed, as are researchers sharing technology across borders. The API carve-out suggests selling access to models is largely unaffected, which limits the commercial bite. The thresholds are broad (the "top-10" test is vague), so real impact depends on how officials apply the rules and on the final catalogue, which is not yet settled.
What is noise
Calling this China's "first" frontier AI export control, and pairing it with US controls, is the framing of one advocacy-style analysis, and it overstates a draft that has not been finalised. MOFCOM itself calls it routine and not aimed at any country, which may be understatement, but nobody has shown intent either way. The source here is a secondary write-up a week after the announcement, so exact thresholds and the immediate effect of the registration rule are unverified against the primary text.
Watch next
- 01The final catalogue after the 15 Nov 2026 comment deadline: whether the 10^26 FLOPs, 2T parameter and top-10 thresholds survive, and whether the AI items are kept, narrowed or dropped.
- 02MOFCOM guidance or case examples on the interim registration platform: approval times, refusals, and whether API/SaaS access stays outside the definition of export.
- 03Reactions from multinationals and Chinese labs, such as changes to model-weight releases, overseas licensing deals or ownership restructures, and any official text confirming the secondary analysis.
Coverage
1 storyNewsletter
Get the no-hype AI news straight to your inbox
Weekly. No spam, unsubscribe any time. Privacy
More regulation signals
Full feed →- NYT-led publishers file summary judgment brief citing internal OpenAI/Microsoft messages calling AI training "astonishing theft" and admitting chatbots substitute for journalism18 Sept 202686
- Anthropic files public S-1 with SEC for Nasdaq IPO under ticker "ANT"26 Oct 202680
- OpenAI discloses sandbox-escape and credential-leak incidents, confirms pause on tool-use for its most capable models26 Sept 202680
- Anthropic threat report: Claude abused for malware, drone/missile software, mass surveillance, and industrial-scale distillation by Chinese AI labs11 Sept 202680