Investment AI, member services, and retirement projections under SPS 530, CPS 230, and the SIS Act.

SPS 530 requires trustees to maintain an investment governance framework that provides effective oversight of all investment activities, including those driven by AI. When algorithms handle portfolio allocation, rebalancing, and asset selection, the board must demonstrate the processes operate within the fund's investment strategy and risk appetite. MySuper product performance is publicly compared through the YourSuper comparison tool, with documented governance required when AI contributes to underperformance.

CPS 230 replaced three previous standards and strengthened operational resilience requirements across all APRA-regulated entities. AI systems supporting investment execution, member account management, and benefit payments must be mapped as critical. Material service provider arrangements covering external investment managers and AI vendors require enhanced due diligence and contractual protections by 1 July 2026. Fourth-party risk is in scope. CPS 234 mandates information security controls and 72-hour incident notification for AI models and training data.

The Financial Accountability Regime extended to superannuation trustees from 15 March 2025. Accountable persons must have clear responsibility for AI systems in their area, with accountability statements that reflect AI obligations. Penalties under FAR can reach $1.565 million for individuals. When an AI-driven investment decision underperforms or a member chatbot crosses the advice boundary, the question is "who is accountable" and "what reasonable steps were taken."

ASIC's "efficiently, honestly and fairly" obligation applies equally to chatbots, retirement calculators, and member nudges. Where AI-generated "general information" ends and "personal advice" begins triggers AFSL obligations. AI-powered calculators must use reasonable assumptions and produce defensible estimates that members rely on. ASIC REP 798 found many licensees lack guidelines for disclosing AI use to consumers, and the December 2026 Privacy Act amendments will require transparency for automated decisions that significantly affect individuals.