Hong Kong Ethical Artificial Intelligence Framework (Version 2.0, December 2025)
Section 4.1.5: System Deployment practices – Hong Kong Ethical Artificial Intelligence Framework (Version 2.0, December 2025)

Hong Kong Ethical Artificial Intelligence Framework (Version 2.0, December 2025) 4.1.5.2: Assume failure: layered mitigations, business rules, output checks and a disclosure document

What does 4.1.5.2 require?

Assuming the application may fail, mitigation is built in to limit damage: multiple layers of mitigation in code to stop harmful outputs, business rules for quality control, checks that identify unexpected, untruthful or inaccurate outputs including through APIs, and a disclosure document made available to users describing the application.

Maintained by Gerard Blokdyk

Other controls in Section 4.1.5: System Deployment practices – Hong Kong Ethical Artificial Intelligence Framework (Version 2.0, December 2025)

Query this from an agent

The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.