Shanghai Artificial Intelligence Laboratory (Shanghai AI Lab)

What it does

State-backed AI research laboratory, a CnAISDA member, and the most substantive frontier-safety research operation in China. Published a Frontier AI Risk Management Framework (v1.0, with Concordia AI) and an accompanying technical risk-analysis report assessing frontier models against cyber-offence, biological/chemical, persuasion, and loss-of-control risk categories — methodologically comparable to Western frontier safety frameworks. Also produced the AI-45° Law / SafeWork line of research on co-evolving safety and capability. Pure research; no regulatory power.

Its powers

research

Honest assessment

This is the one Chinese institution where a foreign AI-safety researcher has a genuinely functioning channel: it publishes in English on arXiv, collaborates openly with Concordia AI, and its researchers respond to academic correspondence. That is a real, working route — but it is a research channel, not a governance or complaint channel. It carries no institutional obligation to respond.

Notes

Its frontier risk framework is the strongest evidence against the claim that China does not engage with catastrophic risk. It explicitly covers loss-of-control and uncontrolled self-replication.