AI Inquiry Atlas

An unofficial, independent guide to the public records of the Joint Select Committee on Artificial Intelligence. Not connected with the Parliament of Australia or the committee. Its makers also made a submission (no. 481; see About). Summaries and scores are written by AI and can be wrong: check the original before relying on them.

Submission 248

Dr Yongxin Xu

Submitter group: Individuals.

Position labels are not shown for individual submitters.

An AI model identified the submitter group and wrote the summary, key points and recommendation summaries. Counts do not measure importance or merit.

Open in the interactive guide

Summary

Read the original submission for the full text.

A finance academic, drawing on four working-paper studies, describes AI as emerging infrastructure in financial markets. It recommends mapping offshore AI provider dependencies, clarifying prudential and advice rules, disclosure and benchmarking for consumer AI financial tools, a broader sovereign capability definition, and loan-level lending reporting.

Key points

Quotes are the submitter’s own words. Text taken from documents, including scanned pages read by text-recognition software, can differ from the PDF. Check the original. Listed in the same order as on the submission’s page in the interactive guide.

  1. Argues that because supply of AI providers is concentrated, dependence on them is potentially a single point of failure in market-wide information processing.

    First, a material share of short-horizon price discovery now runs through these LLM systems, so when they fail, public information takes longer to reach prices. Second, because supply is concentrated, this is potentially a single point of failure in market-wide information processing rather than a set of independent firm-level risks.

  2. Argues CPS 230’s entity-by-entity approach cannot see shared provider dependencies across regulated entities.

    APRA’s Prudential Standard CPS 230, in force since 1 July 2025, requires regulated entities to identify critical operations, manage material service providers and plan for substitutability. The architecture is right, but three gaps remain. It operates entity by entity and therefore cannot see that many entities depend on the same provider.

  3. Argues that general-purpose AI assistants used for investment guidance give output weaker than professional advice, in ways users cannot detect.

    These tools are free, fluent and confident, however their output is systematically weaker than the professional advice it displaces, in ways a user cannot detect from the answer.

  4. Argues scam-focused protection does not reach this problem, and that general-purpose tools are not clearly giving personal advice under existing regulation.

    This is not fraud, so consumer protection work aimed at scams and deception does not reach it. Nor is it clearly personal advice under existing financial advice regulation, since the tool is general-purpose and the user chose to ask it.

  5. Argues that sovereign capability should include assured access to capable models and capacity to evaluate how they handle Australian information.

    Define sovereign AI capability to include assured access and independent evaluation capacity, not domestic model development alone, and fund Australian capability to evaluate how models perform on Australian information.

  6. States that Australia lacks the loan-level reporting needed to determine whether AI adoption in lending expands or restricts credit.

    This means it is not possible for anyone, including a regulator, to determine whether AI adoption in Australian lending is expanding or restricting access to credit, or for whom.

Recommendations

Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.

  1. Ask the Council of Financial Regulators to map and publicly report where financial institutions, market infrastructure and participants depend on a small number of offshore AI providers, including dependencies embedded in third-party software.

    Recommendation 1. Ask the Council of Financial Regulators to map, and report publicly on, where Australian financial institutions, market infrastructure and market participants depend on a small number of offshore AI providers, including dependencies that arrive embedded in third-party software.

  2. Confirm through APRA guidance that generative AI providers fall within CPS 230 material service provider requirements where they support critical operations, and that substitutability planning includes operating without a specific model provider.

    Recommendation 2. Confirm through APRA guidance that generative AI providers fall within the material service provider requirements of CPS 230 where they support critical operations, and that substitutability planning includes the ability to operate without a specific model provider.

  3. Require consumer-facing AI tools answering questions about specific securities or financial products to disclose within the response whether live retrieval was used and the date of the most recent source relied upon.

    Recommendation 3. Require consumer-facing AI tools that answer questions about specific securities or financial products to disclose, within the response, whether the answer used live retrieval and the date of the most recent source relied upon.

  4. Ask ASIC to publish guidance on when a general-purpose AI assistant offering security-specific suggestions engages the financial product advice regime.

    Recommendation 4. Ask ASIC to publish guidance on when a general-purpose AI assistant offering security- specific suggestions engages the financial product advice regime.

  5. Fund periodic independent benchmarking of consumer-facing AI financial outputs, with results published, on the model of ASIC shadow shopping reviews.

    Recommendation 5. Fund periodic independent benchmarking of consumer-facing AI financial outputs, with results published, on the model of ASIC shadow shopping reviews.

  6. Define sovereign AI capability to include assured access and independent evaluation capacity, not domestic model development alone, and fund Australian capability to evaluate model performance on Australian information.

    Recommendation 6. Define sovereign AI capability to include assured access and independent evaluation capacity, not domestic model development alone, and fund Australian capability to evaluate how models perform on Australian information.

  7. Introduce loan-level lending outcome reporting sufficient to audit whether AI adoption expands or restricts credit access, and for whom.

    Recommendation 7. Introduce loan-level lending outcome reporting sufficient to audit whether AI adoption expands or restricts credit access, and for whom.

Read the original

Open in the interactive guide