AI governance in Australian manufacturing.

A person conducting a business or undertaking must ensure, so far as is reasonably practicable, the health and safety of workers, including where AI controls robots, autonomous plant or production lines. Risks from AI-driven machinery must first be eliminated, then minimised, with adequate information, training and supervision.

Designers, manufacturers, importers and suppliers of plant must ensure it is without risks so far as is reasonably practicable and must pass safety information down the supply chain. AI-enabled or robotic plant needs testing, hazard analysis and clear guidance for the next user in the chain.

The regulations set prescriptive rules for managing plant risks, guarding, control systems and hazardous chemicals. Where AI adjusts machine behaviour or chemical processes, the operator must maintain effective control measures, labelling and safety data, and register or license high-risk plant where required.

Manufacturers face strict liability for goods with a safety defect, meaning safety not what persons generally are entitled to expect. AI used in design, quality assurance or automated inspection does not shift this liability, so defective outputs reaching consumers can expose the manufacturer to compensation claims.

A supplier that becomes aware of a death, serious injury or illness associated with use or foreseeable misuse of consumer goods must notify the Commonwealth Minister via the ACCC within two days. Failure to report is a strict-liability offence, so incidents linked to AI-controlled products must be escalated promptly.

Manufacturers handling worker, customer or connected-product data must comply with the Australian Privacy Principles covering collection, use, disclosure and security. NB: the new automated-decision transparency duty in APP 1.7 commences 10 December 2026, requiring privacy policies to disclose computer-based decisions that significantly affect individuals.

The ten guardrails set expected practice for governance, risk assessment, data quality, testing, human oversight, transparency and record keeping. Although voluntary, they signal the direction of proposed mandatory guardrails for high-risk settings and are a practical baseline for factory automation and predictive systems.

Entities with annual consolidated revenue of at least 100 million dollars must publish an annual statement describing modern slavery risks in operations and supply chains and the actions taken. Manufacturers using AI for supplier screening must still address labour risks across global component and material supply chains.

Map every AI or autonomous system on the production line and run a WHS risk assessment before deployment, recording controls and residual risk

Build AI outputs into product-safety testing and quality assurance so defective goods are caught before supply under the ACL manufacturer-liability regime

Stand up a two-day incident escalation process so any serious injury linked to a product triggers ACL section 131 reporting to the ACCC

Update privacy policies and automated-decision disclosures ahead of the 10 December 2026 transparency duty for decisions affecting individuals

Adopt the Voluntary AI Safety Standard guardrails as an internal baseline, assigning accountability, testing and human oversight for each AI system

Ensure suppliers of AI-enabled plant pass down safety information and that imported machinery meets Australian WHS design and supply duties

Check whether food-and-grocery assets trigger the CIRMP obligation administered by CISC; a critical defence industry asset is SOCI critical infrastructure but is not among the 13 asset classes prescribed in s 4(1) of the CIRMP Rules