AI governance in Australian construction.

A person conducting a business or undertaking must ensure, so far as is reasonably practicable, the health and safety of workers and others. Where AI drives fatigue detection, hazard prediction or safety monitoring on site, the tool forms part of the risk controls and the duty holder remains accountable for its reliability. Poor AI outputs that lead to injury can found an offence under the Act.

The model WHS Regulations require a Safe Work Method Statement for each of the eighteen categories of high risk construction work before it starts (reg 291, 299). AI planning, scheduling or estimation tools that reorder tasks or crews must not undermine the identified control measures, and the PCBU must keep the SWMS current.

Duty holders must eliminate or minimise risks from plant so far as is reasonably practicable. AI-enabled or semi-autonomous plant such as automated cranes, machine-control dozers and site robots must be risk assessed, with workers given information, training and supervision and a clear human override before use.

Biometric data used for site access or facial time-and-attendance is sensitive information that generally needs consent before collection, as confirmed in Lee v Superior Wood [2019] FWCFB 2946. Handling of worker and subcontractor personal information must meet the Australian Privacy Principles. NB: the Privacy Act automated-decision transparency duty commences 10 December 2026.

Camera, computer and tracking surveillance of workers must be overt, with at least fourteen days written notice, and the definition of employee reaches volunteers and labour-hire workers as well as direct employees. AI-powered CCTV analytics, drone monitoring and GPS tracking fall within these rules, and covert surveillance is an offence unless authorised by a covert surveillance authority issued by a Judge of the Local Court.

The National AI Centre released the Guidance for AI Adoption on 21 October 2025, evolving the Voluntary AI Safety Standard into six essential practices for governance, risk, transparency, testing and human oversight. It is voluntary and complements existing law, and is a practical baseline for AI safety, estimation and inspection tools.

Each state and territory has security-of-payment legislation giving contractors a right to progress payments on a pay-now-argue-later basis, enforced through statutory adjudication. Where AI automates payment-claim assessment or scheduling, firms must still meet the statutory timeframes and issue valid payment schedules.

Since 9 November 2023, proposing or relying on unfair terms in standard-form small business or consumer contracts is prohibited and carries substantial penalties. Standard-form subcontracts and AI or software supply agreements are caught, and terms allowing one party to unilaterally vary AI-driven pricing or scope may be challenged.

Map every AI tool used for site safety, estimation, scheduling and quality inspection against the WHS primary duty, and confirm a competent person reviews AI outputs before they drive controls

Keep Safe Work Method Statements current and ensure AI planning or optimisation tools never override identified high-risk construction work controls

Risk assess AI-enabled and autonomous plant, and provide workers information, training and supervision plus a clear human override before deployment

Obtain valid consent for biometric site access and audit worker data handling against the Australian Privacy Principles ahead of the 10 December 2026 automated-decision transparency duty

Give overt written notice at least fourteen days before any camera, drone, GPS or AI analytics surveillance, and extend that notice to subcontractors

Review standard-form subcontracts and AI software agreements to remove unfair contract terms, and confirm automated payment tools meet security-of-payment timeframes

Adopt the National AI Centre Guidance for AI Adoption practices as a documented governance baseline for all deployed AI systems