The Hong Kong University of Science and Technology · FACULTY OF ARTS & HUMANITIES

MAIE5103 Chap.10 AI Governance Frameworks and Release Decisions

- one subject, every graph, every model, every mark
8 Chapters4-page Bible
Our own words - no uploaded lecturer files
Updated for this semester
Chapter 10 of 14 · MAIE5103

AI Governance Frameworks and Release Decisions

Four regimes, four kinds of authority

The lectures compare four instruments, and the first rule is to cite each with its jurisdiction and date.

The NIST AI Risk Management Framework 1.0, of 26 January 2023, is voluntary, with a generative AI profile added in July 2024. The EU AI Act entered into force on 1 August 2024 as binding regulation organised by risk tier, with many duties applying from 2 August 2026. Mainland China's interim measures for generative AI took effect on 15 August 2023 and cover filing, security, content and labelling for public services.

Hong Kong, as of the lecture, has no dedicated AI Act and works through the Personal Data (Privacy) Ordinance and the privacy commissioner's guidance.

Govern, map, measure, manage

The NIST framework organises risk work into four functions. Govern sets owners, authority, risk tolerance, resources and accountability, and cuts across the rest.

Map covers purpose and non-uses, the people affected and the ways the system could fail. Measure covers valid targets and tests, group and system evidence and uncertainty. Manage covers mitigation, monitoring, response and the decision to pause, roll back or retire. The lecture stresses that these are iterative functions, not a checklist or a certification.

Hong Kong's 2024 model framework has four parts of its own and recommends levels of human involvement matched to risk.

Documents and the release decision

Datasheets, model cards and decision logs matter when they change a review, a decision or a handover.

They feed a release decision with three outcomes: go for a bounded use with monitoring, a limited pilot to resolve a stated uncertainty, or no-go where the purpose cannot be justified or a harm cannot be controlled. Running a pilot never suspends privacy or accountability duties, or the need for a safe fallback.

In this chapter

What this chapter covers

  • 01

    Pairing every rule with its jurisdiction and date

  • 02

    The NIST AI Risk Management Framework and its generative profile

  • 03

    The EU AI Act's risk tiers and staged duties

  • 04

    Mainland China's interim measures for generative AI

  • 05

    Hong Kong's PDPO and the privacy commissioner's model framework

  • 06

    Govern, map, measure and manage

  • 07

    Datasheets, model cards and decision logs as interfaces

  • 08

    Go, limited pilot or no-go

Worked example · free

Correct three governance claims in a draft report

Q [6 marks]. AskSia-authored practice. A group's draft report contains three sentences: (1) 'NIST requires all US companies to test their AI for bias.' (2) 'The EU AI Act has applied in full since 2024.' (3) 'In Hong Kong, the PCPD framework is the AI law.' Correct each sentence so it matches the lectures. The marks shown are an AskSia study allocation, not a University marking scheme.
The instruments the lectures cite, placed on their stated dates.The instruments the lectures cite, placed on their stated dates. The voluntary NIST framework came first, binding measures in Mainland China and the EU followed, and many EU duties apply from August 2026.2023202420252026NIST AI RMF 1.0China interim measuresPCPD model frameworkNIST generative profileEU AI ActMany EU duties apply
The instruments the lectures cite, placed on their stated dates. The voluntary NIST framework came first, binding measures in Mainland China and the EU followed, and many EU duties apply from August 2026.
  • 2Correct the claim about NIST.
  • 2Correct the claim about the EU AI Act.
  • 2Correct the claim about Hong Kong.
First, the NIST AI Risk Management Framework is voluntary: it describes four functions, including measurement of fairness and bias, but it requires nothing of anyone. Second, the EU AI Act entered into force on 1 August 2024 with staged application, and the fairness lecture notes that many of its duties apply from 2 August 2026, so 'in full since 2024' is wrong. Third, Hong Kong had no dedicated AI Act as of the lecture; the privacy commissioner's 2024 model framework is guidance, its recommendations are not legislation, and duties under the Personal Data (Privacy) Ordinance continue to apply alongside it.
Sia tip — Before writing 'must' about any AI rule, check two things: is the instrument binding, and has the duty you are citing started to apply?
Glossary

Key terms

AI Risk Management Framework
NIST's voluntary 2023 framework that organises AI risk work into the Govern, Map, Measure and Manage functions.
Risk Tier
A category in the EU AI Act that determines which duties apply to a system, from prohibited practices to high-risk obligations.
Limited Pilot
A release decision that allows bounded use in order to resolve a stated uncertainty, while privacy duties, accountability and a safe fallback all still apply.
Human-in-Command
The level of human involvement Hong Kong's model framework suggests where neither human-in-the-loop nor human-out-of-the-loop fits.
Decision Log
A record of why a release was approved, which trade-offs were accepted, who owns monitoring and what triggered any rollback.
Non-use
A use a system is explicitly not intended for, mapped before release so that misuse can be recognised and refused.
FAQ

AI Governance Frameworks and Release Decisions FAQ

What are the four functions of the NIST AI Risk Management Framework?

Govern, Map, Measure and Manage. Govern covers owners, authority, risk tolerance, resources and accountability and runs across the others. Map covers purpose, non-uses, stakeholders and failure paths. Measure covers valid targets, tests, evidence and uncertainty. Manage covers mitigation, monitoring, response and the decision to pause, roll back or retire.

The framework is voluntary and the functions are iterative, not a checklist.

Does Hong Kong have an AI law?

Not a dedicated one, as of the lecture. AI systems that handle personal data fall under the Personal Data (Privacy) Ordinance, and the privacy commissioner has issued ethical AI guidance in 2021, a model personal data protection framework in 2024, and agentic AI privacy guidance noted for August 2026. The framework's recommendations are guidance rather than legislation.

How do I decide whether an AI system is ready to release?

Work through the lecture's release questions: is the purpose legitimate and was a non-AI alternative compared, do the targets and tests fit the actual use, are residual harms, appeals and owners explicit, and are monitoring and rollback funded and tested? Then choose go for a bounded use, a limited pilot to resolve a named uncertainty, or no-go. The lecture presents this as a teaching tool, not an official decision tree.

Study strategy

Assessment move

Build a four-row table of the regimes with jurisdiction, date, binding or voluntary, and one thing each asks of a deployer, and test yourself on it until the dates come without effort. Then practise the release decision on one system per week, writing the three outcomes and which you chose. Governance answers lose marks mainly through wrong dates and 'must' attached to guidance, and both are fixed by the table.

Working through AI Governance Frameworks and Release Decisions in MAIE5103? Sia is AskSia’s AI Arts and Humanities tutor — ask any MAIE5103 AI Governance Frameworks and Release Decisions question and get a clear, step-by-step explanation grounded in how MAIE5103 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.

A+Everything unlocked
Unlocks this Bible + all 13 of your subjects at The Hong Kong University of Science and Technology, plus 1,000+ more Bibles across the AskSia library.
Sia - your MAIE5103 tutor, unlimited, worked the way the exam marks it
The full 4-page Bible + practice bank with worked solutions
Chrome extension - sync your LMS so Sia knows your deadlines
Bilingual EN / Chinese on every Bible and every Sia answer
$0.99 Trial
30-day money-back · cancel in one tap · how it works