artificial-intelligence.giGlobal AI Monitor

Hong Kong AI Regulation & Governance

HK10 active modules2 quiet

Last updated · 12 modules · 6 instruments · 39 claims · 48 sources in the cumulative register

Framework status
The Digital Policy Office's Ethical AI Framework is non-binding guidance, originally for government use and now promoted more broadly, with no statutory force.
Main regulator

RegulatorDigital Policy Office (DPO, Innovation, Technology and Industry Bureau) — Ethical AI Framework & HK Generative AI Technical and Application Guideline

Next milestone
No upcoming dated milestone recorded
Record status
Live · Periodic active record · 48 sources (40 retrieved) · 0 under review · Not yet expert-reviewed
AI-assembled from cited sources
Evidence-gated · How we verify ·
DisclosureNot yet expert-reviewed · Not legal, compliance or investment advice. Facts are machine-assembled from cited public sources. Individual claims have not each been independently verified by a human reviewer. See methodology.

Source tiers: T1 enacted statute, regulation, or binding supervisory order · T2 official regulator guidance, consultation, or registry · T3 international standard-setter or peer-reviewed analysis · T4 practitioner commentary or secondary reporting. Confidence: high = corroborated by retrieved T1/T2 text · medium = single official source or secondary corroboration · low = preliminary or unverified. Full definitions →

Research checked · Updated
Record history

Research as of · Record last changed · Published · Page built

Jump to module (12)

#

The Digital Policy Office's Ethical AI Framework is non-binding guidance, originally for government use and now promoted more broadly, with no statutory force.

0 instruments (0 current) · 0 obligations · 8 claims · 0 milestones

Standing sub-brief · as of 12 Nov 2026 198 words · last cycle 11 Nov 2026 · AI research, not expert reviewed

Hong Kong has no dedicated horizontal AI statute in force; artificial intelligence is governed through sectoral law and non-binding regulator guidance. The jurisdiction's core governance instruments are voluntary and carry no statutory force. The Digital Policy Office's Ethical AI Framework was originally developed to assist government bureaux and departments and is now promoted more broadly, remaining non-binding with no statutory force. The Digital Policy Office's Generative AI Technical and Application Guideline, updated in December 2025, is non-binding guidance addressed to developers, service providers and users of generative AI, recommending watermarking or labelling of AI-generated content. The Office of the Privacy Commissioner for Personal Data's AI Model Personal Data Protection Framework, issued 11 June 2024, is non-binding guidance recommending that organisations using AI systems establish AI governance steering committees.

Looking ahead, Hong Kong's 2026 Policy Address announced a new Commissioner for AI post, AI agent safety guidelines targeted for 2027, a review of deepfake law and a review of AI product liability, none of which has yet been enacted. A cross-government Working Group, steered by the Secretary for Justice, is reviewing Hong Kong law for gaps impeding wider AI application, following a Steering Committee meeting on 6 March 2026.

Claims and evidence (8)

  • The Digital Policy Office's Ethical AI Framework is non-binding guidance, originally for government use and now promoted more broadly, with no statutory force.

    Confidence: Confirmed · As of 1 Dec 2025 · Evidenced
    Evidence and sources

  • The DPO's Generative AI Technical and Application Guideline, updated December 2025, is non-binding guidance for developers, service providers and users of generative AI, recommending watermarking/labelling of AI-generated content.

    Confidence: Confirmed · As of 1 Dec 2025 · Evidenced
    Evidence and sources
    • Digital Policy Office, HKSAR Government T1 · Retrieved · retrieved 2 Nov 2026The Digital Policy Office (DPO) of the Government of the Hong Kong Special Administrative Region (the HKSAR Government) has commissioned the Hong Kong
    • HKSAR Government T1 · Retrieved · retrieved 2 Nov 2026service users should explicitly indicate by watermarks, labels, metadata, digital signatures, etc, whether generative AI has been involved

  • The PCPD's AI Model Personal Data Protection Framework, issued 11 June 2024, is non-binding guidance recommending AI governance steering committees for organisations using AI systems.

    Confidence: Probable (capped: best source is T3) · As of 11 Jun 2024 · Evidenced
    Evidence and sources
    • Mayer Brown T3 · Retrieved · retrieved 2 Nov 2026on 11 June 2024, the Hong Kong Office of the Privacy Commissioner for Personal Data (PCPD) joined them by issuing the "Artificial Intelligence: Model Personal Data Protection Framework"

  • Hong Kong's 2026 Policy Address announced a new Commissioner for AI post, AI agent safety guidelines targeted for 2027, a deepfake-law review and an AI product-liability review, none yet enacted.

    Confidence: Probable · As of 16 Sep 2026 · Evidenced
    Evidence and sources
    • The Standard T2 · Retrieved · retrieved 2 Nov 2026Hong Kong will combat the use of AI in criminal activities, including deepfakes and AI-generated fraud, through legislation
    • Digital Watch Observatory T2 · Retrieved · retrieved 2 Nov 2026a society-wide set of AI agent safety-management guidelines is planned to be released in 2027

  • A cross-government Working Group, steered by the Secretary for Justice, is reviewing Hong Kong law for gaps impeding wider AI application, following a 6 March 2026 Steering Committee meeting.

    Confidence: Confirmed · As of 6 Mar 2026 · Evidenced
    Evidence and sources

  • As at the run date, Hong Kong has no dedicated horizontal AI statute; AI use is governed through sectoral law and non-binding regulator guidance.

    Confidence: Probable (capped: best source is T3) · Evidenced
    Evidence and sources
    • CMS T3 · Retrieved · retrieved 2 Nov 2026There is currently no dedicated AI law in force in Hong Kong.

  • Hong Kong has no dedicated horizontal AI statute in force; AI is governed through sectoral law and non-binding regulator guidance.

    Confidence: Probable (capped: best source is T3) · Evidenced
    Evidence and sources
    • CMS T3 · Retrieved · retrieved 2 Nov 2026There is currently no dedicated AI law in force in Hong Kong.
    • Herbert Smith Freehills Kramer T3 · Surfaced · retrieved 11 Nov 2026Hong Kong does not currently have a dedicated artificial intelligence regulation.
    • AI Policy Tracker T3 · Retrieved · retrieved 2 Nov 2026There is currently no dedicated AI law in force in Hong Kong.

  • Hong Kong's core AI governance instruments (Ethical AI Framework, Generative AI Guideline, PCPD Model Framework) are voluntary guidance with no binding statutory effect.

    Confidence: Probable (capped: best source is T3) · As of 28 Sep 2026 · Evidenced
    Evidence and sources
    • Herbert Smith Freehills Kramer T3 · Surfaced · retrieved 11 Nov 2026Hong Kong does not currently have a dedicated artificial intelligence regulation.

#

Nothing recorded for this module in the research of 2026-11-11.

0 instruments · 0 obligations · 0 claims

#

HKMA Supervisory Guidance for Responsible Adoption of Artificial Intelligence (ART Framework) (Valid from 2026-10-15).

3 instruments (2 current) · 2 obligations · 5 claims · 0 milestones

Standing sub-brief · as of 12 Nov 2026 202 words · last cycle 11 Nov 2026 · AI research, not expert reviewed

Binding obligations for AI used in high-risk use cases currently rest on the Securities and Futures Commission's circular to licensed corporations on the use of generative AI language models (SFC Circular 24EC55), in effect from 12 November 2024. The SFC treats the use of an AI language model to generate investment recommendations, advice or research for clients as a high-risk use case, triggering extra risk-mitigation duties for licensed corporations. Under that treatment, licensed corporations must validate the accuracy of outputs, apply human review, test robustness and disclose AI use for high-risk use cases, and must notify the SFC of new or materially changed high-risk use cases. Licensed corporations must also maintain senior-management oversight and, for in-house AI models, validate, test and continuously monitor them.

Instruments (3)
  • HKMA Supervisory Guidance for Responsible Adoption of Artificial Intelligence (ART Framework) Valid from 2026-10-15
    Regulatory expectation
  • SFC Circular to licensed corporations - Use of generative AI language models Valid from 2024-11-12
    Binding
  • HKMA Circular - High-level Principles on Artificial Intelligence Ended 2026-10-15 Not current
    Regulatory expectation

Obligations

  • para 18-19

    LCs using AI language models for investment recommendations, advice or research must apply high-risk-use-case risk-mitigation duties.

    Risk management · Applies to: Other · SFC Circular to licensed corporations - Use of generative AI language models
  • Principles 1-9

    Authorized institutions must apply nine principles across Accountability, Resilience and Trust throughout the AI lifecycle.

    Risk management · Applies to: Other · HKMA Supervisory Guidance for Responsible Adoption of Artificial Intelligence (ART Framework)
Claims and evidence (5)

  • The SFC treats AI-language-model-generated investment recommendations, advice or research as a high-risk use case for licensed corporations.

    Confidence: Probable · As of 12 Nov 2024 · Evidenced
    Evidence and sources
    • Securities and Futures Commission T1 · Retrieved · retrieved 2 Nov 2026The requirements of this circular apply to LCs offering services or functionality
    • Legal 500 / A&O Shearman T3 · Retrieved · retrieved 2 Nov 2026Notably, the SFC considers using AI LMs for providing investment recommendations, investment advice

  • SFC-licensed corporations must maintain senior-management oversight and, for in-house AI models, validate, test and continuously monitor them.

    Confidence: Probable (capped: best source is T3) · As of 12 Nov 2024 · Evidenced
    Evidence and sources
    • Freshfields T3 · Retrieved · retrieved 2 Nov 2026licensed corporations engaging in in-house AI model development or customisation must

  • LCs must validate accuracy, apply human review, test robustness and disclose AI use for high-risk AI use cases, and must notify the SFC of new or materially changed high-risk use cases.

    Confidence: Probable (capped: best source is T3) · As of 12 Nov 2024 · Evidenced
    Evidence and sources
    • BABL AI T3 · Surfaced · retrieved 2 Nov 2026LCs must notify the regulator when they introduce or materially change high-risk AI use cases.

  • The HKMA's 2019 High-level Principles on Artificial Intelligence, which required bank boards and senior management to remain accountable for AI application outcomes, were superseded on 15 October 2026 by the ART Framework.

    Confidence: Confirmed · As of 15 Oct 2026 · Evidenced
    Evidence and sources
    • Hong Kong Monetary Authority T1 · Retrieved · retrieved 2 Nov 2026Banks are expected to take these principles into account when designing and adopting their
    • AI Sherpa T3 · Retrieved · retrieved 2 Nov 2026Annex A replaces the HKMA's 2019 High-level Principles on AI.

  • From 15 October 2026 the HKMA requires authorized institutions to apply nine principles across Accountability, Resilience and Trust throughout the AI lifecycle, superseding the 2019 High-level Principles.

    Confidence: Probable · As of 15 Oct 2026 · Evidenced
    Evidence and sources
    • Hubbis T2 · Retrieved · retrieved 2 Nov 2026The Hong Kong Monetary Authority (HKMA) announced on 15 October the "ART" Framework, standing for Accountability, Resilience and Trust
    • AI Sherpa T3 · Retrieved · retrieved 2 Nov 2026Annex A replaces the HKMA's 2019 High-level Principles on AI.

#

Nothing recorded for this module in the research of 2026-11-11.

0 instruments · 0 obligations · 0 claims

#

HKMA Circular - Consumer Protection in respect of Use of Generative Artificial Intelligence (Valid from 2024-08-19).

3 instruments (3 current) · 0 obligations · 7 claims · 0 milestones

Standing sub-brief · as of 12 Nov 2026 224 words · last cycle 11 Nov 2026 · AI research, not expert reviewed

Binding disclosure duties currently derive from the Securities and Futures Commission's circular to licensed corporations on generative AI language models (SFC Circular 24EC55, in effect from 12 November 2024), under which licensed corporations using AI for high-risk use cases must disclose AI use to clients alongside validating accuracy, applying human review and testing robustness. Authorized institutions with significant Hong Kong operations must extend the Hong Kong Monetary Authority's 2019 guiding principles to customer-facing generative AI applications on a risk-based approach. From 15 October 2026 the HKMA's ART Framework requires authorized institutions to apply nine principles across Accountability, Resilience and Trust throughout the AI lifecycle.

Instruments (3)
  • HKMA Circular - Consumer Protection in respect of Use of Generative Artificial Intelligence Valid from 2024-08-19
    Regulatory expectation
  • HKMA Supervisory Guidance for Responsible Adoption of Artificial Intelligence (ART Framework) Valid from 2026-10-15
    Regulatory expectation
  • SFC Circular to licensed corporations - Use of generative AI language models Valid from 2024-11-12
    Binding
Claims and evidence (7)

  • The DPO's Generative AI Technical and Application Guideline, updated December 2025, is non-binding guidance for developers, service providers and users of generative AI, recommending watermarking/labelling of AI-generated content.

    Confidence: Confirmed · As of 1 Dec 2025 · Evidenced
    Evidence and sources
    • Digital Policy Office, HKSAR Government T1 · Retrieved · retrieved 2 Nov 2026The Digital Policy Office (DPO) of the Government of the Hong Kong Special Administrative Region (the HKSAR Government) has commissioned the Hong Kong
    • HKSAR Government T1 · Retrieved · retrieved 2 Nov 2026service users should explicitly indicate by watermarks, labels, metadata, digital signatures, etc, whether generative AI has been involved

  • A deepfake pornography incident at HKU prompted a PCPD criminal investigation and official acknowledgement that no specific offence targets AI-generated indecent images.

    Confidence: Probable · As of 29 Jan 2026 · Evidenced
    Evidence and sources
    • Hong Kong Free Press T2 · Retrieved · retrieved 2 Nov 2026there is currently no specific offence targeting the production of indecent photos of other people

  • LCs must validate accuracy, apply human review, test robustness and disclose AI use for high-risk AI use cases, and must notify the SFC of new or materially changed high-risk use cases.

    Confidence: Probable (capped: best source is T3) · As of 12 Nov 2024 · Evidenced
    Evidence and sources
    • BABL AI T3 · Surfaced · retrieved 2 Nov 2026LCs must notify the regulator when they introduce or materially change high-risk AI use cases.

  • From 15 October 2026 the HKMA requires authorized institutions to apply nine principles across Accountability, Resilience and Trust throughout the AI lifecycle, superseding the 2019 High-level Principles.

    Confidence: Probable · As of 15 Oct 2026 · Evidenced
    Evidence and sources
    • Hubbis T2 · Retrieved · retrieved 2 Nov 2026The Hong Kong Monetary Authority (HKMA) announced on 15 October the "ART" Framework, standing for Accountability, Resilience and Trust
    • AI Sherpa T3 · Retrieved · retrieved 2 Nov 2026Annex A replaces the HKMA's 2019 High-level Principles on AI.

  • Authorized institutions with significant Hong Kong operations must extend the HKMA's 2019 guiding principles to customer-facing generative AI applications on a risk-based approach.

    Confidence: Confirmed · As of 19 Aug 2024 · Evidenced
    Evidence and sources
    • Hong Kong Monetary Authority T1 · Retrieved · retrieved 2 Nov 2026focusing on four major areas, namely governance and accountability, fairness, transparency and disclosure, and data privacy and protection

  • The PCPD's August 2021 Guidance sets seven non-binding Ethical Principles for AI development and use.

    Confidence: Probable (capped: best source is T3) · As of 18 Aug 2021 · Evidenced
    Evidence and sources
    • Chambers and Partners T3 · Retrieved · retrieved 2 Nov 2026the AI Guidance and the Model Framework do not impose mandatory requirements and their recommendations are not exhaustive

  • The DPO's Generative AI Technical and Application Guideline is non-binding guidance recommending disclosure (via watermarks, labels, metadata or digital signatures) of AI involvement in content or decisions.

    Confidence: Probable · As of 1 Dec 2025 · Evidenced
    Evidence and sources
    • Digital Policy Office, HKSAR Government T1 · Retrieved · retrieved 2 Nov 2026The Digital Policy Office (DPO) of the Government of the Hong Kong Special Administrative Region (the HKSAR Government) has commissioned the Hong Kong
    • HKSAR Government T1 · Retrieved · retrieved 2 Nov 2026service users should explicitly indicate by watermarks, labels, metadata, digital signatures, etc, whether generative AI has been involved

#

Hong Kong's Copyright Ordinance deems the person who arranged the creation of a computer-generated work (including AI-generated works) to be its author and default copyright owner.

0 instruments (0 current) · 0 obligations · 2 claims · 0 milestones

Standing sub-brief · as of 12 Nov 2026 62 words · last cycle 11 Nov 2026 · AI research, not expert reviewed

Hong Kong's Copyright Ordinance deems the person who arranged the creation of a computer-generated work, including an AI-generated work, to be its author and default copyright owner, a provision currently in force. Following its 2024 copyright-and-AI consultation, the Hong Kong government concluded that existing technology-neutral copyright infringement rules should continue to apply to AI training, rather than introducing a dedicated text-and-data-mining exception.

Cited evidence in this module rests solely on tier 3/4 commentary despite primary official sources existing for this jurisdiction.

Claims and evidence (2)

  • Hong Kong's Copyright Ordinance deems the person who arranged the creation of a computer-generated work (including AI-generated works) to be its author and default copyright owner.

    Confidence: Probable (capped: best source is T3) · Evidenced
    Evidence and sources
    • Lexology T3 · Surfaced · retrieved 2 Nov 2026the author is taken to be the person by whom the arrangements necessary for the creation of the work

  • Hong Kong's government concluded, after its 2024 copyright-and-AI consultation, that existing technology-neutral copyright infringement rules should apply to AI training rather than adding a dedicated TDM exception.

    Confidence: Probable (capped: best source is T3) · Evidenced
    Evidence and sources
    • Lexology T3 · Surfaced · retrieved 2 Nov 2026the Government considers that infringement provisions of the CO should remain general, broad and technology-neutral

#

HKMA Circular - Use of Artificial Intelligence for Monitoring of Suspicious Activities (Valid from 2024-09-09).

1 instrument (1 current) · 2 obligations · 13 claims · 0 milestones

Standing sub-brief · as of 12 Nov 2026 310 words · last cycle 11 Nov 2026 · AI research, not expert reviewed

Across financial-sector regulators, binding and supervisory-expectation instruments now govern AI use. The Securities and Futures Commission's circular on generative AI language models (SFC Circular 24EC55) is mandatory for licensed corporations and effective from 12 November 2024. The Hong Kong Monetary Authority's circular of 9 September 2024 required authorized institutions with significant Hong Kong operations to submit an AI-for-AML-monitoring implementation plan by 31 March 2025. The HKMA updated its Supervisory Policy Manual on capacity building on 28 October 2024 to set AI-related training expectations for banks, with no equivalent general obligation outside banking. Authorized institutions with significant Hong Kong operations must also extend the HKMA's 2019 guiding principles to customer-facing generative AI applications on a risk-based approach. The Insurance Authority has no AI-specific guideline; its Guideline on Cybersecurity, GL20, treats AI as a factor within insurers' cyber-risk profile.

Instruments (1)
  • HKMA Circular - Use of Artificial Intelligence for Monitoring of Suspicious Activities Valid from 2024-09-09
    Regulatory expectation

Obligations

  • Implementation Requirements for AIs

    Affected authorized institutions must submit an AI-for-AML-monitoring implementation plan to the HKMA by 31 March 2025.

    Documentation · Applies to: Other · HKMA Circular - Use of Artificial Intelligence for Monitoring of Suspicious Activities · Deadline
  • para 1

    Licensed corporations using generative AI language models must maintain human-in-the-loop review of all AI model output before any outcome is actioned.

    Human oversight · Applies to: Provider
Claims and evidence (13)

  • The HKMA updated its Supervisory Policy Manual on capacity building on 28 October 2024 to set AI-related training expectations for banks, with no equivalent general obligation outside banking.

    Confidence: Confirmed · As of 28 Oct 2024 · Evidenced
    Evidence and sources

  • The SFC's 12 November 2024 circular on generative-AI language models is mandatory for licensed corporations and effective immediately.

    Confidence: Probable (capped: best source is T3) · As of 12 Nov 2024 · Evidenced
    Evidence and sources
    • HKSOKA T3 · Retrieved · retrieved 2 Nov 2026issued the Circular to Licensed Corporations on the Use of Generative AI Language Models (Ref 24EC55), which is mandatory

  • Authorized institutions with significant Hong Kong operations must extend the HKMA's 2019 guiding principles to customer-facing generative AI applications on a risk-based approach.

    Confidence: Confirmed · As of 19 Aug 2024 · Evidenced
    Evidence and sources
    • Hong Kong Monetary Authority T1 · Retrieved · retrieved 2 Nov 2026focusing on four major areas, namely governance and accountability, fairness, transparency and disclosure, and data privacy and protection

  • The PCPD's June 2024 Model Framework recommends, without legally mandating, that organisations procuring or using AI apply existing PDPO data protection principles.

    Confidence: Probable (capped: best source is T3) · As of 11 Jun 2024 · Evidenced
    Evidence and sources
    • Mayer Brown T3 · Retrieved · retrieved 2 Nov 2026on 11 June 2024, the Hong Kong Office of the Privacy Commissioner for Personal Data (PCPD) joined them by issuing the "Artificial Intelligence: Model Personal Data Protection Framework"
    • DataGuidance T3 · Retrieved · retrieved 2 Nov 2026On June 11, 2024, the Office of the Privacy Commissioner for Personal Data (PCPD) published the Artificial Intelligence: Model Personal Data Protection Framework

  • The PCPD's August 2021 Guidance sets seven non-binding Ethical Principles for AI development and use.

    Confidence: Probable (capped: best source is T3) · As of 18 Aug 2021 · Evidenced
    Evidence and sources
    • Chambers and Partners T3 · Retrieved · retrieved 2 Nov 2026the AI Guidance and the Model Framework do not impose mandatory requirements and their recommendations are not exhaustive

  • The PCPD published non-binding agentic-AI guidance on 25 August 2026 setting nine recommendations covering data minimisation, retention and purpose limitation.

    Confidence: Probable (capped: best source is T3) · As of 25 Aug 2026 · Evidenced
    Evidence and sources
    • Reed Smith T3 · Retrieved · retrieved 2 Nov 2026Published on 25 August 2026, this is the first guidance from the PCPD specifically addressing the privacy risks
    • Dentons T3 · Surfaced · retrieved 2 Nov 2026On 25 August 2026, the Office of the Privacy Commissioner for Personal Data ('PCPD') published Artificial Intelligence: Model Personal Data Protection Framework - Protecting Personal Data Privacy in the Use of Agentic AI ('Guidance').

  • The HKMA's 9 September 2024 circular required authorized institutions with significant Hong Kong operations to submit an AI-for-AML-monitoring implementation plan by 31 March 2025.

    Confidence: Probable (capped: best source is T3) · As of 9 Sep 2024 · Evidenced
    Evidence and sources
    • Timothy Loh LLP T3 · Retrieved · retrieved 2 Nov 2026On 09 Sep 2024, the Hong Kong Monetary Authority (HKMA) issued a circular encouraging Authorized Institutions (AIs) to adopt

  • The Insurance Authority has no AI-specific guideline; its Guideline on Cybersecurity (GL20) treats AI as a factor in insurers' cyber-risk profile.

    Confidence: Probable · Evidenced
    Evidence and sources
    • Insurance Authority T1 · Surfaced · retrieved 2 Nov 2026This Guideline is issued by the Insurance Authority ("IA") pursuant to section 133 of the Insurance Ordinance (Cap. 41)

  • The Digital Policy Office's Ethical AI Framework is non-binding guidance, originally for government bodies and now a voluntary reference for any organisation, built on a three-lines-of-defence governance model.

    Confidence: Probable · As of 1 Dec 2025 · Evidenced
    Evidence and sources
    • Digital Policy Office, HKSAR Government T1 · Retrieved · retrieved 2 Nov 2026The Ethical AI Framework was originally developed to assist government bureaux/departments
    • Lexology T3 · Surfaced · retrieved 2 Nov 2026In December 2025, the blueprint was promoted more broadly as a reference for any organisation operating AI systems in Hong Kong
    • Digital Policy Office T1 · Retrieved · retrieved 2 Nov 2026The Government has formulated the "Ethical AI Framework" in 2021 to provide guidelines when implementing projects

  • The DPO's Ethical AI Framework requires government bureaux/departments to operate a three-lines-of-defence governance model for AI projects.

    Confidence: Probable (capped: best source is T3) · As of 22 Jun 2026 · Evidenced
    Evidence and sources
    • Law.asia T3 · Retrieved · retrieved 2 Nov 2026the DPO released the Hong Kong Generative Artificial Intelligence Technical and Application Guideline (updated in December 2025)

  • The Hong Kong Judiciary's July 2024 guidelines permit judges, judicial officers and support staff prudent GenAI use under seven guiding principles, excluding delegation of judicial functions and judgment-writing.

    Confidence: Probable · Evidenced
    Evidence and sources
    • Hong Kong Judiciary T1 · Retrieved · retrieved 2 Nov 2026JJOs and the Judiciary's support staff may make prudent and responsible use of generative AI
    • Debevoise & Plimpton T3 · Retrieved · retrieved 2 Nov 2026JJOs and support staff of the Judiciary are required to, among other things, only use devices provided by the Judiciary

  • Outside the financial-regulator circulars, Hong Kong has no sector-specific statute governing automated decision-making in employment, consumer credit or housing.

    Confidence: Probable (capped: best source is T3) · Evidenced
    Evidence and sources
    • Herbert Smith Freehills Kramer T3 · Surfaced · retrieved 11 Nov 2026Hong Kong does not currently have a dedicated artificial intelligence regulation.

  • SFC Circular 24EC55 sets expectations for licensed corporations' use of generative AI language models, effective 12 November 2024.

    Confidence: Probable (capped: best source is T3) · As of 12 Nov 2024 · Evidenced
    Evidence and sources
    • Lexology T3 · Surfaced · retrieved 2 Nov 2026the Government considers that infringement provisions of the CO should remain general, broad and technology-neutral

#

SFC Circular to licensed corporations - Use of generative AI language models (Valid from 2024-11-12).

1 instrument (1 current) · 0 obligations · 2 claims · 0 milestones

Standing sub-brief · as of 12 Nov 2026 85 words · last cycle 11 Nov 2026 · AI research, not expert reviewed

Safety-related duties for AI systems in Hong Kong currently sit within the Securities and Futures Commission's circular to licensed corporations on generative AI language models (SFC Circular 24EC55), binding and in effect from 12 November 2024. Licensed corporations must validate accuracy, apply human review, test robustness and disclose AI use for high-risk use cases, and must notify the SFC of new or materially changed high-risk use cases. Licensed corporations must also maintain senior-management oversight and, for in-house AI models, validate, test and continuously monitor them.

Instruments (1)
  • SFC Circular to licensed corporations - Use of generative AI language models Valid from 2024-11-12
    Binding
Claims and evidence (2)

  • SFC-licensed corporations must maintain senior-management oversight and, for in-house AI models, validate, test and continuously monitor them.

    Confidence: Probable (capped: best source is T3) · As of 12 Nov 2024 · Evidenced
    Evidence and sources
    • Freshfields T3 · Retrieved · retrieved 2 Nov 2026licensed corporations engaging in in-house AI model development or customisation must

  • LCs must validate accuracy, apply human review, test robustness and disclose AI use for high-risk AI use cases, and must notify the SFC of new or materially changed high-risk use cases.

    Confidence: Probable (capped: best source is T3) · As of 12 Nov 2024 · Evidenced
    Evidence and sources
    • BABL AI T3 · Surfaced · retrieved 2 Nov 2026LCs must notify the regulator when they introduce or materially change high-risk AI use cases.

#

SFC-licensed corporations must maintain senior-management oversight and, for in-house AI models, validate, test and continuously monitor them.

0 instruments (0 current) · 0 obligations · 1 claim · 0 milestones

Standing sub-brief · as of 12 Nov 2026 45 words · last cycle 11 Nov 2026 · AI research, not expert reviewed

The only standards-adjacent requirement currently recorded for Hong Kong sits within the Securities and Futures Commission's circular to licensed corporations on generative AI language models (SFC Circular 24EC55): licensed corporations must maintain senior-management oversight and, for in-house AI models, validate, test and continuously monitor them.

Cited evidence in this module rests solely on tier 3/4 commentary despite primary official sources existing for this jurisdiction.

Claims and evidence (1)

  • SFC-licensed corporations must maintain senior-management oversight and, for in-house AI models, validate, test and continuously monitor them.

    Confidence: Probable (capped: best source is T3) · As of 12 Nov 2024 · Evidenced
    Evidence and sources
    • Freshfields T3 · Retrieved · retrieved 2 Nov 2026licensed corporations engaging in in-house AI model development or customisation must

#

The PCPD published findings of compliance checks on 60 organisations' AI use on 19 May 2026, finding no PDPO contravention but flagging agentic AI as an elevated risk area.

0 instruments (0 current) · 0 obligations · 4 claims · 0 milestones

Standing sub-brief · as of 12 Nov 2026 141 words · last cycle 11 Nov 2026 · AI research, not expert reviewed

The Securities and Futures Commission's circular of 12 November 2024 on generative AI language models is mandatory for licensed corporations and effective immediately. The Office of the Privacy Commissioner for Personal Data published findings on 19 May 2026 of compliance checks carried out on 60 organisations' use of AI, finding no contravention of the Personal Data (Privacy) Ordinance but flagging agentic AI as an elevated risk area. The Privacy Commissioner for Personal Data has also opened a criminal investigation following a deepfake pornography incident at the University of Hong Kong, with the authority acknowledging that no specific offence currently targets AI-generated indecent images.

Hong Kong's 2026 Policy Address announced a new Commissioner for AI post, AI agent safety guidelines targeted for 2027, a review of deepfake law and a review of AI product liability; none of these has yet been enacted.

Claims and evidence (4)

  • The PCPD published findings of compliance checks on 60 organisations' AI use on 19 May 2026, finding no PDPO contravention but flagging agentic AI as an elevated risk area.

    Confidence: Probable · As of 19 May 2026 · Evidenced
    Evidence and sources

  • Hong Kong's 2026 Policy Address announced a new Commissioner for AI post, AI agent safety guidelines targeted for 2027, a deepfake-law review and an AI product-liability review, none yet enacted.

    Confidence: Probable · As of 16 Sep 2026 · Evidenced
    Evidence and sources
    • The Standard T2 · Retrieved · retrieved 2 Nov 2026Hong Kong will combat the use of AI in criminal activities, including deepfakes and AI-generated fraud, through legislation
    • Digital Watch Observatory T2 · Retrieved · retrieved 2 Nov 2026a society-wide set of AI agent safety-management guidelines is planned to be released in 2027

  • The SFC's 12 November 2024 circular on generative-AI language models is mandatory for licensed corporations and effective immediately.

    Confidence: Probable (capped: best source is T3) · As of 12 Nov 2024 · Evidenced
    Evidence and sources
    • HKSOKA T3 · Retrieved · retrieved 2 Nov 2026issued the Circular to Licensed Corporations on the Use of Generative AI Language Models (Ref 24EC55), which is mandatory

  • A deepfake pornography incident at HKU prompted a PCPD criminal investigation and official acknowledgement that no specific offence targets AI-generated indecent images.

    Confidence: Probable · As of 29 Jan 2026 · Evidenced
    Evidence and sources
    • Hong Kong Free Press T2 · Retrieved · retrieved 2 Nov 2026there is currently no specific offence targeting the production of indecent photos of other people

#

Joint Circular on the Expansion of Generative Artificial Intelligence Sandbox (Sandbox++) (Valid from 2026-03-05).

1 instrument (1 current) · 0 obligations · 8 claims · 0 milestones

Standing sub-brief · as of 12 Nov 2026 237 words · last cycle 11 Nov 2026 · AI research, not expert reviewed

The Hong Kong Monetary Authority, Securities and Futures Commission, Insurance Authority and Mandatory Provident Fund Schemes Authority jointly operate, under the Joint Circular on the Expansion of the Generative Artificial Intelligence Sandbox ('Sandbox++'), a regulatory sandbox in effect from 5 March 2026 for regulated entities to test generative-AI and agentic-AI solutions; applications for the relevant round remained open until 30 June 2026. The original bank-only HKMA GenA.I. Sandbox, launched in 2024, has been superseded by this cross-sectoral Sandbox++, with the superseding joint circular separately stated as effective from 9 September 2026. The expanded Sandbox++ selected 36 agentic-AI use cases for its first cohort, announced 27 August 2026, spanning the banking, securities, asset and wealth management, insurance, Mandatory Provident Fund and stored value facility sectors.

Separately, the Office of the Privacy Commissioner for Personal Data operated an AI privacy sandbox open for applications until 30 October 2026, offering guidance and technical advice to selected applicants.

Instruments (1)
  • Joint Circular on the Expansion of Generative Artificial Intelligence Sandbox (Sandbox++) Valid from 2026-03-05
    Regulatory expectation
Claims and evidence (8)

  • The PCPD's AI privacy sandbox was open for applications until 30 October 2026, offering guidance and technical advice to selected applicants.

    Confidence: Probable · As of 25 Aug 2026 · Evidenced
    Evidence and sources

  • The HKMA's expanded GenA.I. Sandbox++ selected 36 agentic-AI use cases for its first cohort, announced 27 August 2026, spanning banking, securities, asset/wealth management, insurance, MPF and stored value facility sectors.

    Confidence: Probable · As of 27 Aug 2026 · Evidenced
    Evidence and sources
    • The Standard T2 · Retrieved · retrieved 2 Nov 202636 use cases have been selected for the first cohort of the GenA.I.

  • The Digital Policy Office's Ethical AI Framework is non-binding guidance, originally for government bodies and now a voluntary reference for any organisation, built on a three-lines-of-defence governance model.

    Confidence: Probable · As of 1 Dec 2025 · Evidenced
    Evidence and sources
    • Digital Policy Office, HKSAR Government T1 · Retrieved · retrieved 2 Nov 2026The Ethical AI Framework was originally developed to assist government bureaux/departments
    • Lexology T3 · Surfaced · retrieved 2 Nov 2026In December 2025, the blueprint was promoted more broadly as a reference for any organisation operating AI systems in Hong Kong
    • Digital Policy Office T1 · Retrieved · retrieved 2 Nov 2026The Government has formulated the "Ethical AI Framework" in 2021 to provide guidelines when implementing projects

  • The DPO's Generative AI Technical and Application Guideline is non-binding guidance recommending disclosure (via watermarks, labels, metadata or digital signatures) of AI involvement in content or decisions.

    Confidence: Probable · As of 1 Dec 2025 · Evidenced
    Evidence and sources
    • Digital Policy Office, HKSAR Government T1 · Retrieved · retrieved 2 Nov 2026The Digital Policy Office (DPO) of the Government of the Hong Kong Special Administrative Region (the HKSAR Government) has commissioned the Hong Kong
    • HKSAR Government T1 · Retrieved · retrieved 2 Nov 2026service users should explicitly indicate by watermarks, labels, metadata, digital signatures, etc, whether generative AI has been involved

  • The DPO's Ethical AI Framework embeds twelve Ethical AI Principles, a four-tier governance model, lifecycle practice guides and an impact assessment template.

    Confidence: Probable (capped: best source is T3) · As of 25 Jul 2024 · Evidenced
    Evidence and sources
    • Regulations.ai T3 · Retrieved · retrieved 2 Nov 2026Formulated by the Digital Policy Office (formerly OGCIO) to embed twelve Ethical AI Principles, a four-tier governance model, lifecycle practice guides, and an AI Application Impact Assessment template

  • From March 2026 the HKMA, SFC, IA and MPFA jointly operate a Sandbox++ for regulated entities to test GenAI/agentic-AI solutions, with applications for the relevant round open until 30 June 2026.

    Confidence: Probable · As of 5 Mar 2026 · Evidenced
    Evidence and sources

  • The HKMA's original GenA.I. Sandbox, launched in 2024, was superseded by Sandbox++ in 2026.

    Confidence: Confirmed · As of 19 Dec 2024 · Evidenced
    Evidence and sources
    • HKSAR Government / HKMA T1 · Retrieved · retrieved 2 Nov 2026The selected participants will be gradually onboarded to a dedicated platform of the Artificial Intelligence Supercomputing Centre

  • The original bank-only HKMA GenA.I. Sandbox was superseded by the cross-sectoral Sandbox++ joint circular, effective 9 September 2026.

    Confidence: Probable (capped: best source is T3) · As of 5 Mar 2026 · Evidenced
    Evidence and sources
    • Lexology T3 · Surfaced · retrieved 2 Nov 2026the author is taken to be the person by whom the arrangements necessary for the creation of the work

#

Cyberport's AI Supercomputing Centre is scaling to 3,000 petaFLOPS in 2025-26 and reported close to 80% utilisation since its December 2024 deployment.

0 instruments (0 current) · 0 obligations · 2 claims · 0 milestones

Standing sub-brief · as of 12 Nov 2026 56 words · last cycle 11 Nov 2026 · AI research, not expert reviewed

Cyberport's AI Supercomputing Centre is scaling to 3,000 petaFLOPS across 2025-26 and has reported utilisation close to 80% since its deployment in December 2024. The Hong Kong Science and Technology Parks Corporation and SenseTime signed a memorandum of understanding on 16 June 2026 to build what is described as Hong Kong's largest domestic AI data centre.

Claims and evidence (2)

  • Cyberport's AI Supercomputing Centre is scaling to 3,000 petaFLOPS in 2025-26 and reported close to 80% utilisation since its December 2024 deployment.

    Confidence: Confirmed · Evidenced
    Evidence and sources
    • InvestHK (HKSAR Government) T1 · Retrieved · retrieved 2 Nov 2026As Hong Kong's largest AI supercomputing centre, AISC is set to enhance computing power to 3,000 petaFlops in 2025, adding to Hong Kong's strengths in basic scientific research.
    • Digital in Asia T4 · Retrieved · retrieved 2 Nov 2026began phased deployment in December 2024 and has averaged close to 80% utilisation since, with Cyberport at points reporting more than 90% of capacity in use

  • HKSTP and SenseTime signed an MoU on 16 June 2026 to build Hong Kong's largest domestic AI data centre.

    Confidence: Probable · As of 16 Jun 2026 · Evidenced
    Evidence and sources
    • Hong Kong Science and Technology Parks Corporation T1 · Retrieved · retrieved 2 Nov 2026Hong Kong Science and Technology Parks Corporation (HKSTP) today signed a Memorandum of Understanding (MoU) with SenseTime to launch a strategic collaboration on artificial intelligence (AI) computing infrastructure and the AI industry ecosystem.
No modules match.

Filters combine as OR inside a group and AND across groups.

Authorities

No authorities recorded in this research.

Enforcement actions

No enforcement actions recorded in this research.

Cases

No cases recorded in this research.

Consultations

No consultations recorded in this research.

What this record does not yet establish is set out on the transparency register; its dated events are on the regulatory calendar.

Think something on this page is wrong? Report an error.

Technical provenance (record schema, run ID & build metadata)
Schema version
artificial-intelligence-record-v2.12
Research run ID
aic-union-HK-2026-11-04
Page model
module-card-page-v1
Renderer
render_jid_aic_v13
Build date