aiqbench analysis
This is a policy clarification about how China's export controls apply to AI large models. The key distinction is that regulation depends on whether restricted technology physically or technically moves out of China, not on whether a product is offered to overseas users via API or SaaS. Readers tracking AI deployment abroad should note that the restricted category is narrow, and that localized overseas deployment of listed models may still require licensing.
MOFCOM: Most AI Technologies Can Still Be Freely Exported After Measures Take Effect
36Kr learned that a spokesperson for the Ministry of Commerce answered reporters' questions on the revision of the Catalogue of Technologies Prohibited and Restricted from Export. Question: After the Catalogue takes effect, do various overseas "go global" approaches for AI large models, such as localized deployment overseas and providing model invocation services to overseas users, fall under the scope of technology export management? Answer: At present, AI large models and application products "going global" take many forms. Whether a specific "going global" scenario falls under technology export management depends mainly on whether restricted technologies are transferred from within China to overseas. The scope of AI technologies now subject to control is precisely and narrowly defined. After the measures take effect, most AI technologies can still be freely exported. For AI technologies included in the restricted category, localized deployment overseas falls under technology export management; if deployment is on overseas self-owned cloud or other scenarios with relatively controllable risks, convenience measures such as automatic licensing may apply. For enterprises using restricted large models to provide external API calls, App/web interaction, SaaS applications, or lightweight local-side deployment, these do not fall under the scope of large model technology export management.
Read the full article at 36Kr
