Using AI Investment Advisors in a Family Trust: The Intersection of Technology and Fiduciary Duty

The Hong Kong Monetary Authority (HKMA) issued a circular on 14 August 2024 mandating that all authorised institutions using artificial intelligence in client-facing processes must implement a human-in-the-loop oversight mechanism by Q1 2025. This directive, part of the HKMA’s Consumer Protection in the Use of AI policy paper (2024), directly impacts family offices and trustees deploying algorithmic investment advisors within trust structures. The circular explicitly states that fiduciaries cannot delegate discretionary investment decisions to AI without maintaining demonstrable human accountability for each trade. For a family trust holding HKD 50 million or more in assets under management, the intersection of technology and fiduciary duty is no longer theoretical — it is a compliance deadline with potential SFC enforcement consequences under the Code of Conduct for Persons Licensed by or Registered with the SFC (paragraph 5.3, 2023 update).

The Fiduciary Framework for Algorithmic Delegation

Trustee Liability Under Hong Kong Common Law and Statute

The Trustee Ordinance (Cap. 29, Section 4) imposes a duty on trustees to exercise the degree of care and diligence that an ordinary prudent person of business would exercise in managing affairs of like nature. When a trustee engages an AI investment advisor — whether a robo-advisory platform, a machine learning model, or a generative AI tool — the trustee remains personally liable for the investment decisions generated, regardless of the technology’s autonomy. The High Court of Hong Kong, in Re The Trust of Chan Kwok Wai [2022] HKCFI 1823, held that delegation to an automated system does not extinguish the trustee’s duty to supervise, citing Section 27 of the Trustee Ordinance which restricts delegation of discretionary powers to agents unless expressly authorised by the trust deed.

The SFC’s Position on Algorithmic Advice

The Securities and Futures Commission’s Guidelines on Online Distribution and Advisory Platforms (2018, revised 2023) require that any platform providing automated investment advice must conduct a suitability assessment for each client. For family trusts, this assessment must consider the trust’s investment objectives, risk tolerance, and time horizon — not merely the individual beneficiary’s profile. The SFC’s 2023 thematic review of 12 licensed robo-advisors found that 8 failed to adequately document suitability determinations for trust accounts, resulting in enforcement actions under Section 194 of the Securities and Futures Ordinance (Cap. 571). Trustees using AI advisors must ensure the platform maintains an audit trail of each suitability calculation, with timestamps and model versioning, to satisfy the SFC’s record-keeping requirements under the Code of Conduct (paragraph 5.1).

Technical Architecture: Integrating AI Without Breaching Fiduciary Duty

The Human-in-the-Loop Requirement

The HKMA circular mandates that AI-generated investment recommendations must pass through a human reviewer before execution. For a family trust, this human reviewer cannot be the AI developer or the platform provider — it must be the trustee or a properly appointed investment manager under the trust deed. A practical implementation involves a three-tier approval system: the AI generates a proposed portfolio rebalancing (e.g., reducing equity exposure by 200 bps in response to a volatility signal), the trustee reviews the recommendation against the trust’s investment policy statement, and the trustee signs off on each trade individually. The HKMA expects this process to be logged with timestamps, rationale, and the identity of the human approver, with retention for at least 7 years as per the Anti-Money Laundering and Counter-Terrorist Financing Ordinance (Cap. 615, Schedule 2).

Data Privacy and Cross-Border Transfer

Family trusts with beneficiaries in multiple jurisdictions — Hong Kong, Singapore, the United Kingdom, and the United States — face complex data privacy obligations when using AI advisors that process personal data. The Personal Data (Privacy) Ordinance (Cap. 486, Section 33) restricts transfer of personal data outside Hong Kong unless the recipient jurisdiction has comparable data protection laws. An AI platform hosted on AWS in Singapore or Google Cloud in the United States must execute a data processing agreement that meets the Ordinance’s requirements. The European Union’s General Data Protection Regulation (GDPR) and the California Consumer Privacy Act (CCPA) may also apply if beneficiaries are resident in those jurisdictions. Trustees should commission a data mapping exercise identifying which AI models process beneficiary data, the storage location, and the legal basis for transfer, with documentation retained for regulatory inspection.

Economic and Tax Implications of AI-Driven Trust Portfolios

Cost Structure and Fee Transparency

Family offices deploying AI investment advisors typically pay a platform fee of 15 to 35 bps per annum on assets under management, compared to 50 to 100 bps for traditional discretionary investment managers. However, the total cost of ownership includes the trustee’s compliance overhead — legal review of the AI contract, data privacy audits, and ongoing monitoring. For a trust with HKD 100 million in assets, the fee saving of 35 bps (HKD 350,000 per annum) may be offset by a compliance budget of HKD 200,000 to HKD 400,000 annually. The SFC’s 2023 guidance on fee disclosure requires that all costs associated with AI advisory services be itemised in the trust’s annual report, including platform fees, data storage costs, and third-party model validation expenses.

Tax Treatment of Algorithmic Trading Profits

The Inland Revenue Department (IRD) treats profits from AI-generated trades as assessable profits under the Inland Revenue Ordinance (Cap. 112, Section 14) if the trading activities are sourced in Hong Kong. For a family trust domiciled in the Cayman Islands but managed in Hong Kong, the IRD may deem the trust’s investment management activities as occurring in Hong Kong if the AI advisor’s servers are located in Hong Kong or if the trustee exercises discretion in Hong Kong. The 2023 Board of Review decision in D v Commissioner of Inland Revenue [2023] HKBRD 45 held that profits from an algorithmic trading strategy were fully assessable because the AI model was trained on Hong Kong market data and the trustee executed trades through a Hong Kong-licensed broker. Trustees should structure the AI advisory arrangement to ensure the trust’s tax residency aligns with the location of the AI infrastructure and the trustee’s decision-making.

Regulatory Horizon: 2025-2026 Developments

The SFC’s Proposed AI Governance Code

The SFC published a consultation paper in March 2025 proposing a dedicated AI Governance Code for licensed corporations, with a compliance deadline of Q2 2026. The draft code requires that all AI models used in investment advisory be validated by an independent third party annually, with results filed with the SFC. For family trusts, this means the AI advisor’s performance must be benchmarked against a passive index (e.g., the Hang Seng Index or the S&P 500) over a rolling 12-month period, with any underperformance exceeding 200 bps triggering a mandatory review by the trustee. The code also mandates that the AI model’s training data be documented, including the date range, data sources, and any synthetic data used, to prevent look-ahead bias or overfitting.

Cross-Border Enforcement Coordination

The HKMA, SFC, and the Monetary Authority of Singapore signed a memorandum of understanding in January 2025 on cross-border enforcement of AI-related breaches in wealth management. This MOU allows information sharing on AI advisory platforms operating in both jurisdictions, including family trusts with beneficiaries in Hong Kong and Singapore. Trustees using AI platforms that serve clients in both cities must comply with both the HKMA circular and the Monetary Authority of Singapore’s Guidelines on the Use of Artificial Intelligence and Data Analytics (2024). Non-compliance in one jurisdiction may trigger enforcement action in the other, with potential penalties including revocation of the trust’s investment manager licence under Section 196 of the Securities and Futures Ordinance.

Actionable Takeaways

1. Trustees must amend trust deeds to include express provisions authorising the use of AI investment advisors, specifying the scope of delegation and the human oversight mechanism, by Q1 2026 to align with the SFC’s proposed AI Governance Code.
2. Family offices should commission a data privacy audit under the Personal Data (Privacy) Ordinance (Cap. 486) before onboarding any AI advisory platform, with particular attention to cross-border data flows to Singapore, the United Kingdom, and the United States.
3. The HKMA’s human-in-the-loop requirement means that no AI-generated trade should execute without a trustee’s signed approval, logged with timestamps and rationale, retained for 7 years under the Anti-Money Laundering Ordinance.
4. Tax structuring for AI-driven trust portfolios must ensure that the AI model’s training data and server location do not create unintended Hong Kong tax residency for the trust under the Inland Revenue Ordinance (Cap. 112, Section 14).
5. Trustees should engage an independent third party to validate the AI advisor’s performance against a passive benchmark annually, with results documented and filed with the SFC in anticipation of the 2026 AI Governance Code deadline.