AI Governance in Australian Construction: WHS Obligations, Project Delivery AI, and Procurement Requirements

Australia's construction sector is adopting AI rapidly — BIM AI, safety monitoring systems, autonomous equipment, and AI-driven project management. The Work Health and Safety Act obligations, procurement requirements, and governance framework for construction companies.

Key Takeaways

Work Health and Safety laws apply to AI safety monitoring systems on construction sites — an AI safety system that fails to detect a safety hazard creates potential liability under WHS legislation for the person conducting a business or undertaking (PCBU).
AI in building information modelling (BIM) and structural design creates professional indemnity risk — AI-generated designs and specifications require the same level of professional sign-off and review as human-generated work.
Commonwealth government construction contracts increasingly include AI governance requirements — contractors working on Commonwealth projects should anticipate AI procurement conditions aligned with the APS AI use policy.
Autonomous construction equipment — autonomous excavators, robotic concrete placement, drone surveying — requires specific safety cases and Safe Work Method Statements under WHS regulations.
AI in construction labour force management — algorithmic scheduling, subcontractor performance scoring, safety compliance monitoring — creates Fair Work Act obligations that parallel those in other sectors.

"Apenas para fins informativos. Este artigo não constitui aconselhamento jurídico, regulatório, financeiro ou profissional. Consulte um especialista qualificado para orientação específica."

AI governance for Australian construction — the 2026 picture

The Australian construction sector is integrating AI rapidly: project planning AI, computer vision for safety monitoring, drone surveying, BIM-integrated AI design, materials and supply chain optimisation, and increasingly AI for workforce management. The legal and regulatory environment shaping how this AI is deployed has shifted significantly in 2025-2026, particularly in NSW, with implications that extend across the industry.

NSW Work Health and Safety Amendment (Digital Work Systems) Act 2026 — the headline change

The most material AI-specific development is the Work Health and Safety Amendment (Digital Work Systems) Act 2026 in NSW — the first Australian jurisdiction to impose specific health and safety duties on employers using AI and algorithmic management tools. The Act establishes that a person conducting a business or undertaking (PCBU) has a primary duty of care to ensure workers are not put at risk by algorithms, artificial intelligence, automation, or online platforms. In practice, this means if task allocation is delegated to an algorithm that creates an impossible workload, responsibility ultimately sits with the employer rather than the algorithm vendor.

The Act is commenced by proclamation — monitor legislation.nsw.gov.au for the commencement order. The NSW coalition opposition has indicated intent to repeal the changes if they form government after the next election, but the legislative direction is established. For construction businesses operating in NSW, this means: digital work systems including AI-driven safety monitoring, productivity tracking, scheduling, and access control fall within WHS scope and must be assessed for risk.

Beyond NSW, the Senate Select Committee on Adopting AI has recommended extending the existing WHS legislative framework to AI workplace risks. The model WHS framework — already adopted in seven jurisdictions (Commonwealth, NSW, Queensland, Northern Territory, ACT, South Australia, Tasmania, Western Australia) — is likely to follow NSW's lead progressively.

WHS Regulation 2025 and psychosocial risks

NSW's Work Health and Safety Regulation 2025 commenced 22 August 2025, replacing the 2017 regulation. Critically, Reg 55C was amended to require psychosocial risks to be managed using the hierarchy of control measures (previously excluded from psychosocial risk management). This is highly relevant for AI in construction: AI-driven scheduling, performance monitoring, and management systems that create unmanageable workloads, surveillance overreach, or job insecurity may constitute psychosocial hazards requiring control under WHS law.

SafeWork NSW issued a prohibition notice pausing a major University of Technology Sydney restructure due to "serious and imminent risk of psychological harm" — demonstrating that regulators are willing to intervene where digital or organisational change creates psychosocial hazards. The same regulatory logic applies to AI deployment in construction businesses.

EU AI Act exposure for Australian construction

For Australian construction companies operating in EU markets or supplying to EU customers, the EU AI Act creates additional obligations. Annex III high-risk categories relevant to construction include:

Safety components in critical infrastructure — AI used in operation of critical infrastructure (water, gas, electricity, road traffic) falls within high-risk.

Employment — AI used for worker monitoring, performance evaluation, recruitment, task allocation classified as high-risk.

Annex I products — AI integrated into regulated products including machinery (Machinery Regulation 2023/1230 effective January 2027) and PPE (effective August 2028).

The May 2026 Digital Omnibus political agreement postponed Annex III standalone high-risk obligations to 2 December 2027 and Annex I (embedded in products) to 2 August 2028. AI literacy obligations under Article 4 have been in force since 2 February 2025.

Building Code and professional duties

The Building Code of Australia (now the National Construction Code) sets the regulatory minimum for building work. AI used in design, simulation, materials specification, or compliance verification must produce outcomes that comply with the NCC. Where AI design output is used in regulatory submissions, the registered building practitioner remains personally accountable — the practitioner cannot defer to AI judgement.

Each state and territory has its own building practitioner registration scheme: NSW Fair Trading building licensing, Queensland Building and Construction Commission (QBCC), Victorian Building Authority (VBA), and equivalent bodies. Licensed practitioners face professional and disciplinary exposure for inadequate supervision of AI-generated work product.

IP and contract issues

Copyright in AI design output. The US Copyright Office's May 2025 guidance — increasingly persuasive in other jurisdictions — confirmed that fully AI-generated works are not protected. For construction businesses, this raises questions about ownership of AI-generated design output, BIM models, drawings, and specifications. Where AI contributes substantially, original human contribution is required for copyright protection.

Professional indemnity insurance. PI insurers in Australia, UK, and elsewhere now expect disclosure of AI use at renewal. Failure to disclose may void coverage. Insurance is also tightening on cover for AI-enabled design errors.

Contract clauses needed. Construction contracts (and subcontracts) increasingly need specific clauses addressing: which AI tools were used in design or analysis; allocation of responsibility for AI output; liability for AI errors; warranty exclusions for AI-influenced work; data protection where AI processes personal data of workers or third parties.

Worker monitoring AI — particular sensitivity

Construction is a sector where worker monitoring AI is commercially attractive: site access control, productivity tracking, fatigue monitoring, PPE compliance via computer vision, location tracking. But this AI deployment creates compounding legal exposure:

Privacy Act. Personal information about workers — including biometric data, location data, behaviour data — is protected. APP 5 (notification), APP 6 (use), APP 11 (security), APP 12-13 (access and correction) apply. The Privacy Act ADM transparency obligation effective 10 December 2026 will require disclosure of AI in significant decisions about workers.

WHS Amendment Act (NSW). Algorithmic management duties apply directly.

Fair Work Act. Adverse action, unfair dismissal, and right-to-disconnect provisions interact with AI-driven workforce management.

Emotion recognition. The EU AI Act (Article 5) prohibits emotion recognition in workplace contexts from February 2025 — an absolute prohibition. Australia has not adopted this directly, but the National AI Centre and Voluntary AI Safety Standard reference EU AI Act expectations.

Voluntary AI Safety Standard

The Australian National AI Centre's Voluntary AI Safety Standard (released September 2024) is the closest Australia has to a domestic AI governance framework. Its ten guardrails align broadly with international expectations and form the practical basis for sector-specific application. Government procurement is beginning to reference the Standard. Enterprise buyers in construction supply chains increasingly request evidence of alignment.

Practical compliance for Australian construction businesses

Map your AI use across the business: design, surveying, productivity monitoring, safety monitoring, scheduling, materials, supply chain. For each, identify which regulatory regimes apply.

For NSW operations: assess existing AI systems against the WHS Amendment (Digital Work Systems) Act requirements. Document risk assessment. Implement controls particularly for AI affecting workload allocation, performance assessment, and worker surveillance. Build psychosocial risk management into AI deployment design.

For EU exposure: classify AI under EU AI Act risk tiers. Confirm AI literacy compliance (Article 4). Prepare for Annex III high-risk obligations from 2 December 2027 and Annex I from 2 August 2028.

Update PI insurance disclosure. Update standard contracts to include AI-specific clauses. Update workforce policies to disclose AI use and provide grievance mechanisms.

For larger construction businesses, implement an AI governance framework (typically aligned with ISO/IEC 42001 or NIST AI RMF) with documented inventory, risk classification, monitoring, and review.

For SMEs, a simpler approach: a one-to-two page AI policy covering approved tools, data handling rules, disclosure requirements, and named accountability is enough to demonstrate due diligence and reduce risk.

Primary sources: SafeWork NSW · Safe Work Australia · National AI Centre