The Artificial Intelligence Association of Hong Kong (AIHK) has introduced voluntary guidelines to help organisations govern their use of AI, with a particular focus on AI agents.

The framework recommends human approval before an agent takes actions that are difficult to reverse, including initiating payments, making contractual commitments or sending external communications.

Titled Guidelines on Responsible Corporate and Institutional Adoption of Artificial Intelligence, the document was developed by AIHK and its Safety & Governance Council, which brings together academics from institutions including the University of Hong Kong (HKU), Hong Kong University of Science and Technology (HKUST), Hong Kong Baptist University (HKBU) and Lingnan University.

AIHK said most organisations have started adopting AI, but governance has struggled to keep pace. Many remain uncertain about which applications require tighter controls and who is accountable when something goes wrong.

This can leave AI confined to pilot projects or lead to wider deployment without adequate safeguards.

“Many organisations are caught between moving too slowly on AI and moving ahead without the right safeguards,”

said Neil Tan, Founder and Chairman of AIHK.

“Our goal is informed, accountable risk-taking, not zero risk, so organisations can scale AI responsibly and meet the expectations of customers and regulators in Hong Kong and overseas.”

Risk tiers determine the level of AI oversight

The guidelines recommend classifying each AI use case into one of three risk tiers, with controls proportionate to its potential impact.

An internal drafting tool, for instance, would require lighter oversight than a system influencing credit, employment or healthcare decisions.

AI agents connected to organisational systems should be classified as elevated risk at a minimum, according to the guidelines.

AIHK also recommends board-level oversight, appointing an executive responsible for AI adoption and maintaining a central register of AI use across the organisation.

The framework covers organisations of different sizes, from listed companies to SMEs, across sectors including financial services, retail, logistics, healthcare and professional services.

It also addresses organisations that procure AI solutions from external vendors rather than develop their own systems.

AIHK emphasises that outsourcing AI procurement does not transfer responsibility for how the technology is used. Organisations should conduct due diligence on suppliers before deploying their solutions.

AI agents require oversight across entire workflows

The dedicated section on AI agents reflects the risks associated with their ability to perform multi-step tasks on behalf of organisations.

During a single workflow, an agent could access sensitive data, interact with third-party software, engage external service providers and initiate transactions.

An error at an early stage could therefore trigger further problems before human intervention.

The guidelines recommend extending oversight beyond the content agents generate to the actions they perform throughout a workflow, with human approval before any action that would be difficult to reverse.

Four-stage roadmap starts with 90-day governance plan

AIHK does not expect organisations to establish mature governance frameworks before adopting AI. Instead, its four-stage roadmap begins with baseline measures that can be implemented within 90 days, including assigning accountability, approving AI tools and defining permitted data use.

Organisations can then progress towards wider AI adoption and independent assurance.

The guidelines also encourage voluntary measures beyond existing requirements, including alignment with ISO/IEC 42001, the international standard for AI management systems, and labelling AI-generated content.

Framework draws on Hong Kong regulations and international practices

The guidelines build on Hong Kong’s existing regulatory and policy framework, including the Personal Data (Privacy) Ordinance and the Digital Policy Office’s Ethical AI Framework.

They also draw on international practices from the EU, the US, the UK, Mainland China, Japan, South Korea and Singapore.

Hong Kong is also advancing its AI adoption agenda.

The government’s 2026 Policy Address outlined plans to deepen AI use, while the Hong Kong Monetary Authority (HKMA) launched Project SAGE at Hong Kong Fintech Week to explore agentic AI applications in the financial sector.

Featured image credit: Edited by Fintech News Hong Kong, based on image by AIHK