China has proposed adding 40 technologies across AI, biotechnology and photovoltaic tech to its catalogue of restricted exports, as it seeks to protect core national assets.

The move comes after Facebook parent Meta Platforms last December acquired China-founded AI agent developer Manus, in a deal that Chinese regulators later ordered it to unwind.

The proposal covers 23 AI technologies, five biomedical technologies and 12 photovoltaic technologies, of which 14 would be prohibited from export and 26 restricted.

The country’s Ministry of Commerce gave as examples proposed prohibitions on AI-driven malicious code generation, certain proprietary cell and gene therapy process platforms, and certain solar cell module production technologies.

Proposed restrictions in the AI field include software vulnerability discovery.

Large models trained on China’s “core data” and then provided as a service would be prohibited from export, while those trained on “important data” and then provided as a service would be restricted.

Large language models

Core large-model technologies, including weights, key checkpoints, source code and architectur, as well as training-optimisation and inference-acceleration techniques, are to be restricted when a model reaches 1026 floating-point operations of compute, 2 trillion parameters, or a “globally leading” capability level, the ministry said.

It defined the latter as a top-10 global ranking at the time of export in mainstream public evaluations.

The government is consulting on the proposed measures through 15 November, but certain interim pre-export registration requirements took effect immediately.