AI is now a WHS issue: What Safe Work Australia’s new guidance means for employers

Artificial intelligence is rapidly becoming part of everyday work.

Organisations are using AI-enabled systems to analyse data, automate routine tasks, monitor equipment, identify patterns and support workplace decisions. Digital technologies such as mobile applications, wearable devices, automated machinery and safety monitoring systems are also becoming increasingly common.

These tools can improve productivity and help organisations manage workplace risks. However, they can also introduce new physical and psychosocial hazards—or change the nature of risks that already exist.

Safe Work Australia has now made the WHS implications clear.

In July 2026, it released new guidance on artificial intelligence and digital technologies in the workplace. The guidance confirms that risks arising from these technologies must be managed under the same WHS framework that applies to other workplace hazards.

For employers, this means AI cannot be treated solely as an information technology, privacy or cybersecurity issue. It must also be considered as part of the organisation’s broader approach to work health and safety.

What does the new guidance cover?

Safe Work Australia’s guidance applies broadly to AI and digital technologies used at work.

This can include:

• AI-enabled software and applications

• automated decision-making systems

• workplace monitoring technology

• wearable devices

• advanced robotics

• autonomous plant and machinery

• virtual and augmented reality

• mobile phones and computers

• digital safety and security systems.

These technologies may be used to automate tasks, monitor worker activity, analyse safety information, predict hazards or support operational decisions.

Safe Work Australia recognises that these systems can create opportunities to improve WHS outcomes. At the same time, poorly designed, selected or implemented technology can expose workers to additional physical or psychological harm.

The question for employers is therefore not simply whether a technology is useful. It is whether the technology can be introduced, used and maintained without creating unacceptable risks.

Existing WHS duties still apply

The emergence of AI does not create a separate category of WHS responsibility.

Under the model WHS laws, a person conducting a business or undertaking must eliminate risks to health and safety so far as is reasonably practicable. Where elimination is not reasonably practicable, risks must be minimised so far as is reasonably practicable.

Safe Work Australia states that risks associated with AI and digital technologies should be managed using the established WHS risk management process.

In consultation with workers and other relevant duty holders, organisations should:

1. identify hazards;

2. assess risks where they are not already known;

3. implement appropriate controls; and

4. monitor and review those controls.

This process should begin before the technology is implemented—not after workers begin experiencing problems.

It should also continue throughout the technology’s lifecycle. Software updates, changes in data inputs, new integrations, expanded use cases and changes in work design may all alter the risk profile.

How can AI create workplace risks?

AI-related risks are not always obvious.

An AI system may not directly injure a worker, but it can influence how work is allocated, monitored, assessed or completed. These changes may increase workloads, reduce worker autonomy, create uncertainty or encourage unsafe reliance on automated outputs.

Several areas deserve particular attention.

Work intensification

Automation is often introduced with the aim of improving efficiency. However, the time saved by technology may be replaced with higher output expectations, shorter deadlines or increased monitoring.

When systems automatically allocate tasks or calculate productivity targets, workers may feel pressure to maintain a pace determined by software rather than by realistic human capability.

This can contribute to excessive job demands, fatigue and psychological strain.

Employers should assess whether the technology genuinely reduces workload or simply increases the volume and pace of work.

Surveillance and loss of trust

Digital systems can monitor location, activity, driving behaviour, computer use, productivity, fatigue indicators and compliance with procedures.

Monitoring may have a legitimate safety purpose. For example, technology could help confirm that a remote worker is safe or that a driver is taking required rest breaks.

However, the same system may create psychosocial risks if workers feel they are being constantly observed or if information collected for safety purposes is later used for performance management.

Safe Work Australia identifies work intensification, surveillance concerns and “scope creep” as potential risks associated with workplace monitoring technologies.

Organisations should clearly define:

• why monitoring is necessary;

• what information will be collected;

• who can access it;

• how it will be used;

• how long it will be retained; and

• whether it will be used for purposes beyond safety.

Transparency is critical to maintaining worker confidence.

Over-reliance on automated decisions

AI can analyse large amounts of information and identify patterns that may be difficult to detect manually.

In a safety context, this could include identifying recurring incident factors, predicting equipment failure or detecting indicators of fatigue.

These capabilities can support better decision-making, but AI outputs should not automatically be treated as correct.

Safe Work Australia warns that over-reliance on technology can itself create risk. An organisation may, for example, rely solely on a fatigue-monitoring system while overlooking a worker’s own report that they are too tired to perform a task safely.

AI systems may produce incomplete, inaccurate or misleading results because of:

• poor-quality data;

• missing context;

• incorrect system configuration;

• biased historical information;

• unexpected workplace conditions; or

• limitations in the system’s design.

Human oversight should therefore remain part of safety-critical decisions.

Workers should also have a clear way to question or override a system output when it conflicts with observable conditions or professional judgement.

Poorly designed automated systems

Automation can remove workers from hazardous environments. Autonomous plant, robotics and remote-controlled equipment may reduce exposure to dangerous machinery, substances or high-risk tasks.

However, automated systems can introduce different hazards.

These may include:

• unexpected equipment movement;

• failures in sensors or interlocks;

• unsafe interaction between workers and machinery;

• inadequate isolation procedures;

• confusion about whether equipment is operating automatically or manually;

• software updates that disable alarms or safety functions; and

• workers entering hazardous areas to troubleshoot faults.

The introduction of automation should be supported by appropriate guarding, isolation, maintenance, competency and emergency procedures.

Employers should not assume that a task has become safe simply because fewer workers are directly involved.

Psychosocial hazards during organisational change

Technology often changes more than a single process. It can alter roles, reporting structures, skill requirements and expectations about how work should be performed.

Workers may be concerned that AI will replace their role, reduce their hours or make their skills less valuable. Others may be expected to use unfamiliar systems without adequate training or support.

Uncertainty, inadequate consultation and poorly managed organisational change can contribute to psychosocial hazards.

These risks may be reduced by involving workers early, clearly explaining how the technology will affect their work and providing sufficient time and resources for training.

Worker consultation is not optional

One of the most important messages in Safe Work Australia’s guidance is the need to consult workers and their representatives.

Consultation should occur while options are still being considered. Asking workers for feedback after a system has already been purchased and configured limits their ability to influence the outcome.

Workers often understand practical hazards that may not be visible to technology vendors, senior management or project teams.

They may be able to identify:

• situations where the system could be bypassed;

• circumstances not reflected in the data;

• tasks requiring human discretion;

• potential conflicts with existing procedures;

• physical or cognitive demands created by the technology;

• barriers affecting workers with different abilities or levels of digital literacy; and

• unintended impacts on workload, autonomy or workplace culture.

Consultation should include health and safety representatives where they are present. Organisations may also need to coordinate with software providers, labour-hire companies, contractors, equipment manufacturers and other businesses that share WHS responsibilities.

A practical WHS checklist before implementing AI

Before introducing an AI or digital system, employers should consider the following questions.

1. What problem are we trying to solve?

The organisation should be able to clearly describe the intended purpose of the technology.

Introducing AI because it is available—or because competitors are using it—is not a sufficient business or safety justification.

2. How will the technology change the work?

Look beyond the technical functionality.

Consider how the system will affect workloads, responsibilities, supervision, worker autonomy, decision-making and interaction between people and equipment.

3. What could go wrong?

Identify credible failure scenarios, including incorrect outputs, system downtime, missing data, unauthorised access, alarm failures and workers relying on the system when it should not be trusted.

4. Who could be affected?

Consider employees, contractors, labour-hire workers, visitors, customers and members of the public.

Different groups may experience different risks.

5. Have workers been consulted?

Consult the workers who will use the system or whose work will be influenced by it.

Record concerns raised and explain how they have been addressed.

6. What controls are required?

Controls may include system design changes, access restrictions, human review, physical safeguards, clear procedures, training, workload limits and escalation processes.

Administrative instructions should not be the only control where a higher-level control is reasonably practicable.

7. What decisions must remain with people?

Clearly define where human authorisation, review or intervention is required, particularly for safety-critical, disciplinary or employment-related decisions.

8. How will effectiveness be monitored?

Establish measures for identifying whether the technology is working safely.

These could include incident reports, worker feedback, system error rates, overdue actions, hazard reports and psychosocial risk indicators.

9. When will the assessment be reviewed?

Reviews should occur after implementation and whenever the system, work environment or intended use changes.

They should also occur following an incident, near miss, worker complaint or unexpected system behaviour.

AI can support WHS—but it cannot own it

AI and digital technologies can make valuable contributions to workplace health and safety.

They can help organisations examine large volumes of safety information, identify recurring patterns, automate repetitive administration and produce stronger first drafts of reports or risk documentation.

Technology can also help make important information easier to find and support more timely follow-up of safety actions.

However, these tools should support—not replace—the people responsible for understanding workplace conditions and making decisions.

An AI-generated risk assessment is still only a draft until it has been reviewed in the context of the actual workplace. A predicted hazard still needs to be investigated. An automatically generated action still needs an appropriate owner, timeframe and verification process.

Responsibility for health and safety remains with the organisation and its duty holders.

What employers should do now

Organisations do not need to wait until they are undertaking a major AI project.

AI features are increasingly being added to software that businesses already use. Employers may also discover that workers are independently using publicly available AI tools to prepare documents, analyse information or support decisions.

A useful starting point is to identify where AI and digital technologies are already being used across the organisation.

From there, employers can:

• review whether existing risk assessments cover the technology;

• determine whether workers were appropriately consulted;

• identify decisions that require human oversight;

• examine potential physical and psychosocial hazards;

• establish clear rules for acceptable use;

• provide appropriate training;

• review how system outputs are checked; and

• monitor whether the technology is creating unintended consequences.

Safe Work Australia’s new guidance does not suggest organisations should avoid AI.

It sends a more practical message: innovation and safety must be considered together.

AI can improve how work is performed and how workplace risks are managed. But those benefits are more likely to be realised when organisations introduce technology deliberately, consult the people affected and apply the same disciplined risk-management approach expected for any other workplace change.

About AI in myosh

myosh is developing practical AI capabilities within the records, forms, dashboards and risk management tools organisations already use.

These capabilities are designed to help users review information more efficiently, identify patterns across safety data and create stronger first drafts of important safety and risk documentation.

AI-generated content should always be reviewed by an appropriately qualified person and considered in the context of the organisation’s actual workplace, legal obligations and risk profile.