紫衡ZiHeng集團Group EN

SFC 講 AI 管治,
你啲工具跟唔跟得上?
SFC is writing AI rules.
Can your tools keep up?

香港證監會近年對「持牌法團用 AI 語言模型」嘅期望越寫越實:問責、人類監督、能力、通知。呢啲唔係新發明——合規嘅本質一向係「講得出依據」。問題係,你手上啲 AI 工具,俾唔俾到依據? Hong Kong's SFC keeps sharpening its expectations for licensed corporations using AI language models: accountability, human oversight, competence, notification. None of this is new — compliance was always "show me the basis". The question is whether your AI tools can produce one.

  1. 問責:答錯咗,邊個孭?Accountability: who owns a wrong answer?

    SFC 嘅邏輯好直接:AI 出嘅力,持牌人孭嘅鑊。用黑箱工具答客,等於將問責交俾一個講唔出理由嘅系統。The SFC's logic is blunt: AI does the work, the licensee carries the can. Answering clients with a black box means outsourcing accountability to a system that can't explain itself.

  2. 監督:AI 唔應該係最後把關嗰個Oversight: AI shouldn't be the last line of defence

    人類監督要可行,先決條件係「有嘢俾人監督」——出處、段落、版本日期,先至係可以覆核嘅物件。得一句答案,監督得個估字。Human oversight needs something to oversee — a source, a paragraph, a version date. A bare answer leaves you supervising a guess.

  3. 能力:識用,仲要識佢唔掂喺邊Competence: know it, and know where it fails

    用 AI 嘅人要知道工具嘅邊界。一個主動講明「我唔係法律意見」、「冇出處我唔答」嘅工具,好過一個乜都答、令你以為佢乜都啱嘅工具。Users must know a tool's boundaries. One that says "not legal advice" and "no source, no answer" beats one that answers everything and lets you assume it's always right.

  4. 通知:高風險用途,自己清楚先Notification: know if your use is high-risk

    SFC 要求高風險 AI 用途通報。第一步唔係填表,係分得清自己嘅用途算唔算高風險——呢個本身已經係一條合規問題。High-risk AI uses must be notified to the SFC. Step one isn't the form — it's knowing whether your use counts. That itself is already a compliance question.

呢套期望,規 Sir 由第一日行到而家RegSir was built to these lines from day one

本文係一般資訊,唔係法律意見。SFC 通函與守則嘅具體要求,以官方原文為準——有疑問,正路係去查原文,或者問規 Sir,等佢指出處俾你。 This article is general information, not legal advice. For the exact requirements in SFC circulars and codes, go to the official text — or ask RegSir and let it show you the source.

三個快問Three quick answers

規 Sir 答得啱,係咪即係合規建議?Is a good RegSir answer compliance advice?

唔係。每條答案都標明「初步合規資訊,非法律意見」——佢俾你出處同原文,重大決策仍然要搵持牌人士。No. Every answer is labelled preliminary compliance information, not legal advice — you get the source and the text; for consequential decisions, see a licensed professional.

SFC 對用 AI 嘅持牌法團有咩期望?What does the SFC expect of LCs using AI?

大致四條線:問責制——高級管理層要對 AI 決策負責;人類監督——AI 唔可以自己話事;能力——用嘅人要識佢嘅限制;通知——高風險用途要向 SFC 通報。Broadly four lines: accountability — senior management owns AI-driven decisions; human oversight — AI doesn't get the last word; competence — users must know its limits; notification — high-risk uses must be reported.

點解「冇出處就冇答案」緊要?Why does "no source, no answer" matter?

因為監管查嘅唔係你答得流唔流利,係你可唔可以指出依據。一條冇出處嘅答案,喺問責制度下等於冇答。Because a regulator doesn't ask how fluent the answer was — it asks what it was based on. An answer without a source is, under an accountability regime, no answer at all.