Skip to main content

    AI strategy and governance

    AI governance for boards (Australia)

    Australian boards oversee AI through the duties they already have. The Australian Securities and Investments Commission (ASIC) has said those duties apply to how an organisation adopts and uses AI. A board needs to know which AI uses are material, who owns each one and how their risks are reported.

    In short

    ASIC says directors' duty of care and diligence extends to AI

    • ASIC's Report 798 (October 2024) places AI within directors' and officers' existing duty of care and diligence.
    • The Human Technology Institute is part of the University of Technology Sydney (UTS). A 2025 survey of 419 Australian corporate leaders by the UTS Human Technology Institute found only about 15% felt confident in their organisation's AI governance.
    • The Board AI Risk Register maps 5 categories of AI risk to the board's oversight. Each category has an owner, controls and a review date.
    • From 10 December 2026, Australian Privacy Principle (APP) 1.7 and APP 1.8 set privacy policy disclosure rules for some automated decisions.

    Detail

    Board AI oversight is part of work the board already does

    ASIC's Report 798, Beware the gap: Governance arrangements in the face of AI innovation, was published in October 2024. It states that directors' and officers' duties of reasonable care and diligence 'extend to the adoption, deployment and use of AI'. For a board, AI oversight is part of work it already does.

    The duty of care and diligence for directors and officers sits in section 180 of the Corporations Act 2001. How that duty applies to a particular AI use is a question for the board's legal advisers. The board can make sure now that it can describe how AI risk is managed across the organisation.

    The Board AI Risk Register is a practical tool for that description. It groups AI risk into 5 categories: strategy, operations, compliance, people and reputation. Where your organisation groups its risks differently, use its categories instead. The register turns each category into a question the board can put to management:

    CategoryQuestion for management
    StrategyWhich AI investments are material to our strategy, and what result is each expected to deliver?
    OperationsWhich of our AI systems act or decide without a person checking the result, and who owns each one?
    ComplianceWhich legal obligations does each material AI use engage, and who assessed them?
    PeopleWhich roles will AI change faster than our people can adapt, and what is our plan for them?
    ReputationHow would customers, staff or regulators learn of an AI failure, and who would respond?

    Each category gets a likelihood rating, an impact rating, existing controls, gaps and an owner. The register sits alongside your existing risk register, so AI risk is managed with your other enterprise risks. The board sets a review cycle, for example quarterly.

    The register also sets which AI decisions need board visibility and which management can handle on its own. A system that makes credit decisions is more likely to need board approval than a customer service chatbot.

    Assume your organisation already has shadow AI, meaning AI used without the organisation's knowledge. Find the AI tools staff already use, then decide which to approve and which to replace.

    From 10 December 2026, APP 1.7 and APP 1.8 set privacy policy disclosure rules for some automated decisions. Our article on what APP 1.7 requires sets out the test in full. Start now if your AI governance framework is yet to address these principles.

    Commercial impact

    Facial recognition determinations applied the Privacy Act

    For organisations the Privacy Act covers, existing law already reaches AI that handles personal information. In October 2024 the Privacy Commissioner determined that Bunnings had breached the Privacy Act in its use of facial recognition technology. In February 2026 the Administrative Review Tribunal upheld the findings on transparency and notification and set aside the finding on collection.

    In August 2025 the Commissioner made a determination against Kmart over its use of facial recognition technology. The determination was published in September 2025. Both matters applied existing Privacy Act obligations to facial recognition technology.

    The Australian Consumer Law applies to what an organisation says and does through AI, for example misleading statements by a chatbot.

    The Privacy and Other Legislation Amendment Act 2024 added APP 1.7 to 1.9 to the Privacy Act 1988. They apply from 10 December 2026. An organisation that starts its review late may have to change its privacy policy before it has finished mapping its decisions.

    AI belongs in the board's risk discussion beside cyber and climate. Deployment is moving quickly, and controls can lag behind it.

    Structured governance also helps if something goes wrong. The Board AI Risk Register records that directors considered AI risks, assessed them against the 5 categories, assigned owners and monitored controls.

    Pilots stall when there is no clear accountability, no risk assessment and no compliance review before deployment. Governance that supplies those 3 things helps successful pilots scale.

    Clear thresholds also speed up low-risk work. When the board has agreed in advance which uses management may approve, those uses do not wait for a board meeting.

    In practice

    Illustrative cases of boards finding unknown AI use

    Illustrative composite: an insurer's board mapped its AI systems in a single workshop. It found AI systems in production that it had not known about. One was processing claims decisions without documented human review. The insurer then set up oversight protocols, retrained the claims team and began reporting on AI to the risk committee each quarter.

    Illustrative composite: a governance review at a mid-market professional services firm found staff using generative AI tools with client data and no contract protections. Some of those tools stored data offshore. The firm had not assessed whether that was a disclosure to an overseas recipient under APP 8. The board then required an organisation-wide AI policy and a list of approved tools with controls to stop sensitive data leaving the firm. It also required a simple AI register covering every deployment.

    Illustrative composite: a large retailer's board had not set up AI governance. Its pricing algorithm produced different prices across regional postcodes, and customers complained. The retailer had no documented risk assessment, no controls framework and no clear owner to respond. Without those records, the complaints were slow and costly to answer.

    Illustrative composite: a financial services board asks for a risk register entry in every AI business case before it approves funding. The step adds a little time to each approval. It also means privacy and other risks are considered before money is committed.

    What to do

    Steps for the board to set AI oversight with management

    • Ask management for a list of every AI system in production, with an owner for each.
    • Put the 5 register questions to management, and set a review cycle for the answers.
    • Agree which AI uses management may approve without coming to the board.
    • Bring AI governance into your existing risk and audit committees, because standalone structures tend to work in isolation.
    • Ask when the privacy policy will be reviewed against APP 1.7 and APP 1.8, and who will sign it off.

    Last reviewed

    Put this topic against your own numbers, owners and timelines.