Skip to main content

    AI risk and compliance

    AI disruption risk: what boards should assess

    AI adoption creates operational, regulatory, reputational and competitive risks. This article helps Australian executives and boards identify what could go wrong. The risks range from AI system failures and Privacy Act breaches to biased results and competitors who move first. It also shows how to separate acceptable experimentation from serious threats, so the board can set its AI risk appetite.

    In short

    Boards should weigh inaction risk beside adoption risk

    • AI is now a material business risk for Australian organisations, yet many boards still treat it as a technology issue.
    • Our 5 categories of AI risk give boards a structured threat model. They are autonomy, misuse for destruction, misuse for power, economic displacement and unknown risks.
    • Boards should weigh the risk of inaction beside the risk of adoption. Adoption mistakes are often recoverable, while lost competitive position can be hard to win back.
    • A Board AI Risk Register turns broad concerns into specific scenarios, each with early warning indicators, an owner and a response plan.

    Detail

    AI risk in 5 categories restated for a single organisation

    AI is now a material business risk for Australian organisations. It is a business risk as well as a technology question, and every board should treat it that way.

    The 5 categories of AI risk give you a structured way to think about what could go wrong. They are adapted from the five risks Anthropic chief executive Dario Amodei sets out in his January 2026 essay 'The Adolescence of Technology'. His list is autonomy, misuse for destruction, misuse for seizing power, economic disruption and indirect effects. The essay describes risks to society, so the framework restates each one for a single organisation.

    Autonomy risk covers AI systems acting beyond their intended boundaries. Misuse for destruction covers bad actors using AI for physical or digital harm. Misuse for power covers a concentration of AI capability that creates unfair advantage or control. Economic displacement covers entry-level white-collar work that may be automated sooner than your workforce plan assumes. Unknown risks are those nobody can predict yet, which is why the board needs governance that can respond to surprises.

    For boards, a Board AI Risk Register is a good place to start. It is a structured document that maps each risk category to your business. For each one, it records your exposure, the early warning indicator, the owner of the response and the mitigation plan.

    The register should cover 4 domains. Competitive risks include AI-enabled competitors taking share, new entrants with AI-native models and your differentiators becoming commodities. Operational risks include AI system failures, vendor outages and poor data quality that produces wrong decisions at scale. Compliance and legal risks include Privacy Act breaches, discrimination by biased AI systems and copyright risk from AI tools that generate content. Reputational risks include customer backlash, employee resistance and media coverage of AI failures.

    Boards should give the risk of inaction the same weight as the risk of adoption. If a competitor deploys AI that materially cuts its costs, your multi-year roadmap can become irrelevant much sooner than you planned.

    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. Those are the people accountable for governance describing their own confidence. Boards should treat that low level of confidence as evidence about the state of governance.

    Commercial impact

    Disruption risk is uneven between adoption and inaction

    The disruption risk is uneven. Getting AI adoption wrong usually costs some capital, some time and some credibility, and the organisation can recover. Getting inaction wrong can cost market position, talent and eventually relevance, which are harder to win back. That difference should shape every board discussion about AI risk.

    The Australian Securities and Investments Commission (ASIC) published Report 798, Beware the gap: Governance arrangements in the face of AI innovation, in October 2024. It states that the duties of directors and officers to act with reasonable care and diligence 'extend to the adoption, deployment and use of AI'. Section 180 of the Corporations Act 2001 sets the duty of care and diligence for directors. A board with no structured response to AI risk has a governance gap, and it should take advice on what its duties require.

    The 5 categories make boards look beyond the obvious risks. Most boards already consider data breaches and compliance, and they should. Fewer have modelled economic displacement, where a large share of entry-level roles changes or disappears. That scenario affects workforce costs, the talent pipeline and the operating model, and it is worth testing now.

    Pilots that start without a baseline, an owner and a decision rule are less likely to reach production. That is a risk in its own right. A board that sees only failed pilots can develop an aversion to AI. It can then miss the value of the pilots that succeed.

    In practice

    Illustrative cases of boards modelling AI disruption

    Illustrative composite: a retailer's board models a competitor deploying AI pricing optimisation. The scenario rests on actual competitive intelligence about the competitor. It shows how far the retailer's margins could move if it did not respond. The board then has a specific risk to weigh against the cost of a response.

    Illustrative composite: a professional services firm tests its operations for economic displacement. A large share of its graduates' work could be automated soon. The firm redesigns its graduate program around work that combines AI with professional judgement, and continues to recruit graduates.

    Illustrative composite: a mid-market logistics company's board sees AI disruption as a problem for large companies only, so it models no competitive AI scenarios. The risk it has not tested is an AI-native entrant using route optimisation and dynamic pricing to win part of its market.

    Illustrative composite: a review of a manufacturer's AI vendors finds that the manufacturer depends on one vendor. Its main AI provider has access to production data, pricing algorithms and supplier information. The contract has no data portability clause, and the company has no exit plan. The board asks for renegotiated terms and a plan to reduce the dependency.

    What to do

    Steps to build and maintain a board AI risk register

    • Build a Board AI Risk Register that maps each of the 5 risk categories to your business.
    • Model both adoption risk and inaction risk in every board discussion, and weigh them honestly.
    • Run scenario planning regularly, including what happens if a competitor deploys AI that materially cuts its costs or your AI vendor fails.
    • Assign an owner and early warning indicators to each category in the register.
    • Review the register on a set cycle, such as quarterly, because AI capabilities, regulation and competition change quickly.

    Last reviewed

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