The Hong Kong securities regulator will write agentic AI guidance

The Securities and Futures Commission of Hong Kong will give more guidance on agentic AI. Agentic AI is software that does many steps without help. The chief executive, Julia Leung Fung-yee, told this at Hong Kong FinTech Week in 2026. A study in July and August found that 30 percent of 20 large firms use agentic AI, and 25 percent are in pilot tests.

What is new

Leung told that a person cannot check each middle step. The guidance will use principles, and it will keep a person in charge of the full workflow.

A cybersecurity framework will set basic requirements for all licensed firms. Firms with more risk will have more requirements.

The numbers

The study included 20 large firms and 78 affiliated licensed corporations in Hong Kong. Leung told that agentic AI can repeat errors in many steps. It can also do tasks that no person approved, and it can cause problems for other firms.

What is not known

The regulator gave no date for the guidance. It did not tell the requirements.

Insight: Requirements for AI agents come after the use

In October 2026, the Hong Kong Monetary Authority released an AI framework for banks. More than 60 percent of the uses in its second sandbox test were agentic AI. On 15 October, Singapore showed a four-phase test handbook for AI agents. Requirements read as a step after the use, and the first document that firms follow can become the standard. Leaders must check which document their regulator uses.