Skip to content

Governance

AI governance questions every Australian lender should ask

By Saby Saxena4 min read

Summarise with

In short

The oversight, privacy, audit and vendor risk questions to answer before AI agents touch borrower data, mapped to ASIC, APRA and the Privacy Act.

AI agents can take real work off lending teams. They can also introduce new risks if they’re rolled out without the right controls. Before any AI tool touches borrower data or customer conversations, lenders need confidence in how it’s governed.

Regulators are paying attention. ASIC’s 2024 report Beware the gap found that some financial services and credit licensees were adopting AI faster than their governance was evolving. APRA’s CPS 230 now sets clear expectations for operational risk and service providers. And privacy obligations around automated decisions are getting tighter.

Below are the questions we think every Australian lender should ask early, whether you’re evaluating a vendor or building in-house. We’ve grouped them so you can share them with risk, compliance, technology and operations.

Oversight and accountability

  • Which actions can the AI take on its own, and which require a person to approve them?
  • Who reviews and signs off on those actions, and do they have the time and information to do it properly?
  • How are unusual, high-risk or vulnerable customer cases escalated?
  • Who in the business is accountable for the AI’s performance and outcomes?
  • How does the AI fit into existing responsible lending and Best Interests Duty processes?

Privacy and data handling

The Privacy Act 1988 and the Australian Privacy Principles apply to borrower data whether a person or an AI system handles it. From 10 December 2026, privacy policies will also need to explain when automated decisions that could significantly affect individuals are used.

  • Where is borrower data stored and processed? Does any of it leave Australia?
  • Is borrower data used to train models, and can that be switched off contractually?
  • How is access controlled, and is every access logged?
  • How long is data kept, and how is it deleted?
  • How will your privacy policy describe any automated decision-making?

Auditability and record keeping

  • Is every AI action recorded with a timestamp, the information it used and the reason for the action?
  • Can you reconstruct what the AI saw and did for a specific borrower if a complaint reaches AFCA?
  • Can records be exported easily for internal audit or a regulator?
  • Are records kept for as long as your obligations require?

Operational resilience and vendor risk

For APRA-regulated lenders, CPS 230 Operational Risk Management has applied since 1 July 2025. If an AI provider supports a critical operation, it may need to be treated as a material service provider. Non-ADI lenders should still expect similar questions from funders and auditors.

  • Is the technology already running in live production with other lenders, or would you be the first?
  • How does it integrate with your origination, core banking and CRM systems?
  • What happens if the service is unavailable? Is there a clear fallback to manual processing?
  • How does the provider meet information security expectations such as those in CPS 234?
  • What are the exit arrangements if you need to change providers?

Performance, testing and fairness

  • How is the AI tested before going live, and against what scenarios?
  • How are errors measured, reported and corrected over time?
  • What does the AI do when it isn’t sure? Does it ask a person or guess?
  • How do you check that outcomes are fair across different borrower groups?
  • Who reviews changes to the underlying models or prompts before they go live?

Customer experience and transparency

  • Do borrowers know when they’re dealing with an AI agent?
  • Can they reach a person easily at any point?
  • How are complaints about AI interactions handled through your internal dispute resolution process?

A useful reference point

The Australian Government’s Voluntary AI Safety Standard, released in 2024, sets out ten guardrails covering accountability, risk management, data governance, testing, human oversight and transparency. It isn’t mandatory, but it’s a handy framework to check your answers against.

We’ve shared our approach on the security and governance page. It isn’t a certification, but it explains how we’re thinking about these questions.

Frequently asked questions

Is there a specific AI law for lenders in Australia?

Not yet. Lenders still need to meet existing obligations, including responsible lending, privacy, consumer protection and, for APRA-regulated entities, prudential standards, when they use AI. Guidance continues to evolve, so check current regulator publications and seek advice.

Who should own AI governance in a lending business?

It works best as a shared responsibility with a clear executive owner, supported by risk, compliance, technology and the operational teams using the AI day to day.

To see how onebank answers these questions on your own scenarios, book a demo.

Saby Saxena

Founder, onebank

See it on your own scenarios

onebank runs the work in between. Your team approves every decision.

More articles

All posts