Taiwan's AI Risk Classification Framework: A different path

On the second day of the U.N.'s "Global Dialogue on AI Governance" in Geneva, 7 July, attracting multistakeholder groups from around the world, the Ministry of Digital Affairs of Taiwan officially announced an "AI Risk Classification Framework" that will serve as a rulebook for all Taiwan government agencies in implementing the AI Basic Act. 

Taiwan's AI Basic Act lays out a unique AI regulatory and governance structure — making it distinct from those adopted in other major jurisdictions, such as the European Union, the United States and other major economies in Asia — and the framework is the core element for visualizing and operationalizing such a system.

A rulebook for the public sector but useful for the private sector, too

First of all, the AI Risk Classification Framework provides general guidance as well as sets of operational steps, workflows and examples for sectoral regulators in Taiwan to assess the potential AI risks that may occur if the businesses they regulate adopt AI technology in their operations. 

It also provides guidance for those regulators to formulate and determine actions to further address risks, for example, issuing self-discipline guidelines, implementing pre-deployment screening or licensing requirements, or proposing the enactment of new statutes or amendments to existing statutes to prohibit or limit the deployment of certain defined high-risk AI. 

Government agencies must follow the principles and methodology set forth within the framework to implement tasks under the AI Basic Act. The private sector, such as AI developers and deployers, is not bound by the framework, but will be bound by future regulations or policies to be implemented based on the framework.