Duologs Observatory — Regulatory Signal
China's AI Safety Governance Framework 3.0 makes agentic risk a distinct governance problem
China released AI Safety Governance Framework 3.0, updating its risk classification and governance approach for recent AI developments. The framework specifically addresses agentic and embodied AI risks, reflecting the shift from systems that primarily answer questions to systems capable of actively performing tasks. The framework is non-binding.
Duologs Lens
The framework is non-binding and should not be represented as regulation. Its importance is directional: governance is beginning to follow AI from model behaviour toward autonomous action.
Board-level question
Could this happen inside your stack?