es. AFS licensees and advisers can use AI to assist in answering client questions, but they remain responsible for the accuracy, appropriateness and compliance of the response. ASIC expects that any client-facing communication, including AI-assisted content, meets advice, disclosure and conduct obligations.
Expanded Answer
AI can be used as a tool to support efficiency, consistency and responsiveness in client communications. However, it does not change the underlying regulatory framework. If an AI-generated response constitutes financial product advice—particularly personal advice to a retail client—all usual obligations apply, including best interests, appropriate advice, and correct disclosures.
AI models such as ChatGPT are best understood as “probabilistic parrots“, a term coined by Sean Graham in May 2025 to describe large language models that generate fluent but non-sentient, non-reasoning output based entirely on probability. It’s not a novel critique but it’s a clearer, more engaging alternative to “stochastic parrots” for general, professional, and regulatory audiences.
These models predict the most statistically likely response, based on their training data—not based on understanding, context, or intent. As such, relying on these “probabilistic parrots” introduces some material risks to your business. Remember, LLMs don’t “know” the law, regulatory frameworks, or client circumstances. They produce output that sounds plausible but may be legally or ethically incorrect. The model’s training data may be outdated, jurisdictionally irrelevant, or biased. Plus, ASIC expects human professionals—not algorithms—to ensure that advice is appropriate, accurate, and tailored. Unreviewed or AI-generated content may breach obligations under RG 175 and s961B of the Corporations Act.
In practice, this means AI outputs must be reviewed, validated and tailored before being provided to clients. Licensees should ensure controls are in place to prevent the provision of inaccurate, incomplete or misleading information, and to distinguish between general information and personal advice. Particular care is needed where AI may “hallucinate” content or provide generic responses that do not reflect the client’s circumstances.
Risk increases when AI is used without oversight, when outputs are relied on without review, or when the boundaries between general and personal advice are unclear. ASIC has indicated that technology use does not reduce accountability and may increase expectations around governance and monitoring. For further context, see our guide for advisers and licensees: navigating AI in financial services and avoid the AI trap: how to stay compliant and ahead of the game.
Why it matters
Incorrect or unreviewed AI responses can lead to misleading advice, breaches of best interests obligations, and client harm. ASIC is increasingly focused on how licensees govern technology use, particularly where it affects client outcomes.
Practical guidance
- Review and approve AI-generated responses before providing them to clients
- Define clear boundaries between general information and personal advice when using AI
- Implement governance, monitoring and training to ensure AI use aligns with compliance systems
Further reading
AI in financial advice: efficiency gains, compliance risks and cognitive costs Y