本文目前仅提供英文版本。
AI Governance in Australian Superannuation: APRA SPG 516, Member Best Interests, and Investment AI
Superannuation funds using AI in investment decisions, member communications, retirement income projections, and claims processing face obligations under SIS Act member best interests duties, APRA prudential standards, and ASIC's conduct framework. The complete governance guide.
Key Takeaways
The SIS Act member best interests duty — requiring fund trustees to act in the best financial interests of members — applies to AI systems used in investment decisions, member communications, and benefit administration.
APRA's Superannuation Prudential Standard SPS 530 (Investment Governance) creates model risk management expectations for AI used in investment strategy, asset allocation, and risk management decisions.
AI-generated retirement income projections and financial advice to members are subject to ASIC's financial services licensing regime — AI advice tools used by funds must satisfy best interests duty and appropriate advice obligations.
Claims processing AI — used in death, total and permanent disability, and income protection claims — creates specific obligations: algorithmic claims decisions must be explainable to members and subject to genuine human review.
The Your Future, Your Super performance test and associated member communications obligations apply to AI-generated performance attribution and member outcome reporting — accuracy and clarity standards are non-negotiable.
"仅供参考。本文不构成法律、监管、财务或专业建议。如需具体指导，请咨询合格专家。"
The member best interests duty and AI
The Superannuation Industry (Supervision) Act's member best interests duty — which requires trustees to act in members' best financial interests — is the foundational governance principle for superannuation AI. Every AI system deployed by a superannuation fund must be assessed against this duty: does this AI system, as deployed, serve the best financial interests of members? This question has specific content for each AI use case.
For investment AI — models used in asset allocation, risk management, manager selection, or factor exposure — the best interests duty requires that the AI produces better investment outcomes for members than available alternatives, that it is properly governed and monitored, and that investment staff understand its outputs and limitations sufficiently to exercise genuine investment judgment. An AI model that investment staff treat as a black box, accepting its outputs without critical evaluation, raises questions about whether the trustee is genuinely exercising its investment governance obligations.
AI in member communications: the advice boundary
Many superannuation funds are deploying AI in member communications — chatbots that answer member enquiries, personalised retirement projections, AI-generated consolidation recommendations, and personalised investment option suggestions. The regulatory challenge is that these communications can cross the boundary from factual information into financial advice, which requires an Australian Financial Services Licence and compliance with the best interests duty and appropriate advice obligations.
The ASIC guidance on scaled and digital advice is directly relevant. AI systems that take into account a member's specific circumstances and make personalised recommendations about financial products or strategies are providing financial advice regardless of whether a human advisor is involved. Funds must assess their AI communication tools against this boundary and ensure that tools that cross into advice territory are governed as advice tools — with appropriate authorisations, compliance frameworks, and member consent arrangements.
Claims processing AI and member rights
AI used in insurance claims processing — assessing death benefit claims, total and permanent disability claims, and income protection claims — affects members at the most difficult moments of their lives. The governance obligations for claims AI are particularly demanding because the stakes are high and because the members affected are often in vulnerable circumstances. ASIC has specifically addressed the obligation of superannuation trustees to handle claims efficiently, honestly, and fairly — an obligation that applies to AI-assisted claims processes as much as to human processes. The key requirements: claims decisions influenced by AI must be explainable to the member in plain language, there must be a genuine human review pathway, and the AI system must be monitored for systematic errors that disadvantage members.