AI and Emerging Technologies

Artificial intelligence and emerging technologies are rapidly changing the way Australian businesses operate. We are using AI to write documents, analyse data, allocate work, schedule workers, monitor performance, identify hazards, assess risk, predict maintenance requirements and increasingly support operational decision-making.

At the same time, automation, wearable technology, advanced robotics, digital monitoring systems and algorithmic decision-making are changing how work is performed.

There are significant opportunities to improve productivity and safety, particularly where technology can remove workers from hazardous environments, identify deteriorating conditions earlier or process large amounts of information that would otherwise be difficult to analyse.

However, businesses also need to consider whether the introduction of new technology creates new risks or changes existing ones.

In July 2026, Safe Work Australia released new guidance specifically addressing artificial intelligence and digital technologies in Australian workplaces. The guidance reinforces that AI and digital technologies do not sit outside existing work health and safety obligations. Where technology creates or changes a risk to workers, those risks need to be managed in the same way as any other workplace risk.

AI Is More Than ChatGPT

When we hear about AI in the workplace, it is easy to think about generative AI tools being used to prepare emails, reports, procedures or presentations.

The issue is much broader.

Safe Work Australia describes digital technologies as including software and applications, mobile devices, wearable technology, security and safety technologies, advanced robotics and automation. Artificial intelligence can include machine learning algorithms and generative AI.

Many organisations may already be relying on digital systems to:

• Allocate jobs and work orders

• Develop employee rosters

• Monitor vehicle and equipment movements

• Predict fatigue

• Monitor worker location

• Assess productivity

• Identify maintenance requirements

• Generate procedures and documentation

• Analyse incidents and trends

• Monitor compliance

• Identify hazards

• Recommend controls

• Operate autonomous or semi-autonomous equipment

• Make or influence decisions about workers

This creates an important shift in risk management. We are no longer only asking whether workers are following the system. We also need to consider whether the system itself is making appropriate decisions and whether those decisions introduce additional risk.

Technology Can Reduce One Risk and Create Another

There is enormous potential for emerging technology to improve workplace safety.

Automation can remove workers from hazardous environments. Sensors can identify hazardous conditions before a worker is exposed. Wearable devices may help identify fatigue indicators. Computer vision systems can potentially identify workers entering exclusion zones while predictive analytics can identify equipment deterioration before failure.

Virtual reality and augmented reality can allow workers to practise high-risk activities without initially being exposed to the actual hazard. AI may also assist organisations to analyse large amounts of incident, hazard, inspection and control verification data to identify patterns that may otherwise be missed.

These are genuine opportunities. However, introducing technology because it reduces one risk does not automatically mean the overall risk has been reduced.

Safe Work Australia specifically identifies that digital technology may improve WHS outcomes while also introducing or changing other risks.

Consider autonomous plant. Removing an operator from a hazardous environment may significantly reduce exposure, however the organisation now needs to consider sensor failure, unexpected movement, interactions between autonomous and manually operated equipment, communication failures, software faults and how workers respond when the system behaves unexpectedly.

The original hazard may have changed, but the need to understand how controls can fail has not.

Automation Can Change the Work Left Behind

One of the more interesting issues identified by Safe Work Australia is work intensification.

We often assume that automation reduces workload. In some situations, it may instead change the type of work being performed.

Consider a worker who previously completed a mixture of routine administrative tasks and complex tasks requiring judgement, concentration and problem solving. If AI automates most of the routine work, the worker may now spend a much greater proportion of their day completing cognitively demanding tasks.

Safe Work Australia identifies this as a potential fatigue and psychosocial risk, particularly where automation results in workers completing a higher proportion of complex work requiring sustained concentration.

This creates a number of questions for organisations to consider.

If technology allows a business to process more work, are performance expectations increased accordingly? If an algorithm continuously allocates tasks, does it account for reasonable workloads and breaks? If productivity systems monitor every aspect of a worker's performance, how does this affect job control, autonomy and trust?

These are not simply technology implementation issues. They are work health and safety issues.

Algorithmic Management Is Becoming a WHS Issue

NSW has taken this issue a step further through the Work Health and Safety Amendment (Digital Work Systems) Act 2026.

The legislation defines a digital work system to include an algorithm, artificial intelligence, automation or an online platform.

Among the amendments is clarification that the existing primary duty of care extends to risks arising from digital work systems. The legislation also introduces a specific duty relating to whether the allocation of work through a digital work system puts workers' health and safety at risk.

SafeWork NSW is currently developing supporting Digital Work Systems Guidelines.

Importantly, as at August 2026, most substantive provisions have not yet commenced. SafeWork NSW has advised that commencement dates are still to be announced and most provisions cannot commence until at least one month after the Digital Work Systems Guidelines are published.

While businesses should not treat these provisions as though they are already fully operational, the direction is clear. Digital systems that influence how people work are increasingly being recognised as part of the work system itself and should be considered when organisations assess how work is designed and managed.

What If AI Gets the Risk Assessment Wrong?

There is another issue that businesses should consider carefully.

AI can produce extremely convincing information. It can also produce information that is incorrect, incomplete or inappropriate for the circumstances.

A professionally presented risk assessment is not necessarily a good risk assessment.

For example, a worker may ask an AI tool to develop a risk assessment for a confined space entry. The system generates hazards, consequences, controls, legislative requirements, emergency arrangements and PPE requirements.

The document looks professional and appears comprehensive.

However, if one of those controls is technically inappropriate for the actual atmosphere, equipment, substances or work environment and nobody identifies the error, the organisation may have created a document that gives workers confidence without providing an effective control.

The issue becomes more significant when AI-generated information is used at scale. An organisation may use AI to generate hundreds of SWMS, procedures, risk assessments, inspection questions or training materials.

A mistake made by one person may affect one document. A mistake built into an automated system may be replicated hundreds or thousands of times.

This means human review and competent technical oversight remain important, particularly where AI-generated information is being used to make decisions that could affect worker health and safety.

Safe Work Australia identifies human oversight and auditing as potential controls to ensure technology continues to operate as intended.

AI as a Critical Control

Organisations should be particularly cautious where AI or digital technology forms part of a critical control.

For example, a mining operation may use technology to identify personnel entering an exclusion zone around mobile equipment. An organisation might reasonably determine that this system forms part of a critical control for a vehicle interaction risk.

If it is a critical control, the organisation should then be able to define exactly what the technology needs to do for the control to be effective.

This may include understanding:

• Detection reliability

• Required response times

• Sensor limitations

• Connectivity requirements

• Alarm functionality

• Software configuration

• System interfaces

• Maintenance requirements

• Failure notifications

• Worker response requirements

• Testing and verification requirements

It is also important to consider how workers interact with the technology. If workers begin to rely heavily on an automated detection system, could this reduce situational awareness or weaken other controls?

The same considerations apply to fatigue monitoring technology, collision avoidance systems, automated isolation systems, AI hazard detection systems and predictive maintenance platforms.

Technology does not remove the need for critical control management. In many cases, it increases the importance of clearly defining control performance requirements and understanding potential failure modes.

Software Updates May Need to Be Considered Through Management of Change

Another interesting issue raised by Safe Work Australia is the impact of changes to existing digital systems.

This includes updates initiated or pushed by software providers.

A digital system may have been assessed when it was originally introduced and may have been operating effectively for months or years. The software provider then releases an update that changes an interface, workflow, alarm threshold, automated decision rule or integration.

The organisation may not have initiated the change, but the change may still affect how work is performed.

This raises an important management of change question.

Should changes to software, algorithms and automated decision-making systems be included within an organisation's management of change process where they have the potential to affect safety-critical activities?

In some cases, the answer is likely to be yes.

As organisations become increasingly dependent on digital systems, understanding how updates are assessed, tested and communicated will become an increasingly important part of operational risk management.

Psychosocial Risks Need to Be Considered

Many risks associated with AI and digital technology will not necessarily arise because the technology fails. They may arise because the technology changes the way work is designed.

Safe Work Australia identifies a range of potential psychosocial hazards associated with digital technologies including high job demands, low job demands, low job control, poor support, lack of role clarity, poor organisational change management, poor organisational justice, inadequate reward and recognition, remote or isolated work, harmful behaviours and fatigue.

An algorithm that continuously allocates work may increase job demands. Digital monitoring systems may reduce worker autonomy. Poorly implemented technology may create additional workload rather than reducing it while automated decision-making may create concerns about fairness or transparency.

Technology may also make workers increasingly accessible outside normal work hours through mobile applications and digital platforms.

The important point is that the system does not need to malfunction to create a WHS risk. It may be operating exactly as it was designed and still create unreasonable workloads, poor job control or other psychosocial hazards.

Worker Monitoring and Surveillance

Digital technology also makes it increasingly easy to monitor workers through GPS tracking, cameras, productivity systems, keystroke monitoring, wearable technology and automated performance analytics.

Some of these systems can provide valuable safety information. For example, GPS tracking may support remote worker safety while wearable technology may help identify potential fatigue or exposure conditions.

However, Safe Work Australia also identifies the risk of scope creep. Technology introduced for one purpose, such as improving safety, may later be used for worker surveillance or performance management.

This can affect worker trust, autonomy and perceptions of organisational justice, particularly where the purpose or use of the information has not been clearly communicated.

Worker consultation therefore becomes particularly important when organisations are introducing technology that monitors behaviour, performance or location.

What Should Organisations Be Doing?

Organisations do not necessarily need to develop a separate AI safety management system.

Existing risk management processes can usually be used, however they may need to be expanded to properly consider the way digital technology changes work.

When introducing or significantly changing AI or another digital technology, organisations should consider:

What Work Is Changing?

Consider how the technology changes tasks, workloads, staffing levels, supervision, decision-making, communication and interactions between workers, plant and systems.

What Hazards Could Be Introduced?

Consider physical hazards, psychosocial hazards, human factors and potential interactions between different hazards.

What Decisions Is the Technology Making?

There is a significant difference between technology providing information to support a competent person's decision and technology making an automated decision without human intervention.

Organisations should clearly define where human judgement and approval are required.

What Happens When the Technology Is Wrong?

Consider foreseeable failure scenarios.

What happens if the data is incorrect? What happens if a sensor fails or connectivity is lost? What happens if the algorithm makes an inappropriate decision or workers misunderstand the information being provided?

Organisations should also consider what happens when a supplier changes the technology.

Are Any Digital Systems Critical Controls?

Where technology prevents or mitigates a serious risk, organisations should define the performance requirements necessary for that control to be effective and establish appropriate testing and verification activities.

Have Workers Been Consulted?

Workers often understand how changes will affect day-to-day work in ways that may not be obvious during system design or procurement.

Consultation should occur early enough for workers to genuinely influence how the technology is introduced and used.

How Will the Risk Be Monitored?

The risk assessment should not finish when the technology is implemented.

Organisations should review incidents, hazards, complaints, workloads, control failures, overrides, alarms and other indicators that may show the technology is creating unintended consequences.

The risk should also be reviewed when the technology changes.

Where Do We Go From Here?

AI and emerging technologies have enormous potential to improve workplace health and safety. They can remove people from hazardous environments, identify weak signals earlier, provide better access to information and analyse data at a scale that would previously have been impossible.

However, organisations should avoid assuming that introducing technology automatically reduces risk.

As technology becomes more involved in allocating work, monitoring performance and making operational decisions, risk assessments will increasingly need to consider not only what could go wrong with the work but also what could go wrong with the systems influencing how that work is performed.

The organisations that manage this well will understand what the technology is intended to achieve, what new risks it introduces, how it can fail and how they will verify that it continues to operate effectively.

Ultimately, the same principle applies whether a control is a physical barrier, a procedure, a worker, a sensor or an algorithm. Organisations still need to be able to demonstrate that the control is effective when it is needed.

Useful Links

• Safe Work Australia: AI and digital technologies covers the broad WHS duties, emerging risks and the fact that software updates can change WHS risk. Safe Work Australia: AI and digital technologies

• Safe Work Australia: Managing risks for sections on work intensification, fatigue, human oversight, psychosocial hazards and reviewing controls when software changes. Managing risks from AI and digital technologies

• Safe Work Australia: Example opportunities and risks for the discussion about technology reducing one risk while introducing another and the examples around automation. Example opportunities and risks

• SafeWork NSW: Digital Work Systems Guidelines. This is important because it confirms that a digital work system includes an algorithm, AI, automation or an online platform and clearly explains which provisions have and have not commenced. SafeWork NSW: Digital Work Systems Guidelines

AI and Emerging Technologies