The image above is confronting because it reminds us what safety systems are meant to protect: a person whose absence will be carried by family, friends and colleagues.

When a safety concern is raised, the first question is not whether it has been entered into a maintenance system. It is whether the hazardous work is still continuing.

Two companies and two directors have been charged with industrial manslaughter following the death of 28-year-old Lachlan Carslake in July 2024, and a project manager has also been charged with a Category 1 offence.[1][2] Court documents reported by ABC News allege that an interlock guarding switch on the milling machine Mr Carslake was operating was in a disabled state and that an external technician had raised safety concerns about the switch two days before the incident.[2]

These are allegations before the court. They have not been tested, and the defendants are entitled to the presumption of innocence. This article does not seek to determine what occurred or assign responsibility.

The issue for every other organisation is not what the court will decide, but what its own system does when credible information shows that a control may not be working. A warning that enters a maintenance system but does not change the operating state is not yet protection. The question, then, is who has the authority, and the obligation, to stop the work.

A reported defect is information, not protection

Most organisations have a process for reporting plant defects. An operator tells a supervisor. A maintenance request is created. The equipment is added to a backlog. Priority and resources are considered. Eventually, the work is completed and the item is closed.

That workflow may be appropriate for an oil leak, a damaged light or a component approaching the end of its service life. It is not automatically appropriate when the defect affects a control relied upon to prevent fatal or serious harm. An interlock, guard, isolation device or emergency stop is not merely another machine component. It changes whether a person can be exposed to the hazardous energy, movement or material the machine creates.

SafeWork SA says guarding must prevent bypassing or disabling, disable plant operation if removed, be properly maintained and control risks from broken or ejected parts and workpieces. Its guidance also says guards must not be overridden or disabled.[3]

In June 2026, SafeWork SA issued a safety alert warning that missing, bypassed or ineffective guarding creates a high risk of entanglement, crushing or amputation. The regulator identified inadequate guarding, poor maintenance, weak isolation procedures and time pressure among common contributing factors.[4]

A critical-control defect is not only a maintenance condition. It is an operational risk state.

The handover where risk disappears

The failure is often not that nobody knew. It is that information moved between systems without ownership of the exposure. The operator reports the issue to the supervisor. The supervisor sends it to maintenance. Maintenance records the defect and waits for parts. Production assumes maintenance is managing it. WHS assumes the line manager has assessed it.

Senior leaders see an ageing work order but not the continued exposure. Every part of the process can appear to have performed its task while the organisation as a whole fails to answer whether the work should continue. This is a classic signal-to-action gap. The warning exists. The system records it. But the information does not change the operating state.

A system that records a failed critical control while allowing the exposure to continue has not managed the risk. It has documented it.

Define what must happen before the control fails

The time to decide what happens when a critical control is unavailable is not after the defect is found. Organisations should define the response in advance.

1. Identify the controls that are genuinely critical

Not every control requires the same response. A critical control is one whose absence or failure would materially increase the risk of a fatal or other major unwanted event. The organisation should be able to identify these controls for its highest-consequence hazards without searching through every line of every risk assessment.

2. Define the impaired state

What evidence shows that the control is unavailable, unreliable, bypassed or outside its operating limits? The trigger may be a failed test, missing guard, disabled interlock, overdue proof test, damaged isolation point, alarm fault or credible report from a worker or technician. Ambiguity at this point is dangerous. If nobody can say when the control is considered failed, the organisation can continue operating while debating terminology.

3. Link impairment to an automatic operating response

What work must stop? What plant must be isolated? What area must be quarantined? If continued operation is ever permitted, what alternative controls are required, who is authorised to approve them and what evidence demonstrates they provide adequate protection?

The response should not depend on whether the person receiving the report is sufficiently assertive to challenge production.

4. Separate repair from permission to return to service

Completing maintenance is not the same as verifying control effectiveness. Before work resumes, a competent person should confirm that the control has been restored, tested under the relevant conditions and returned to its required performance standard. Closing the work order should follow verification, not substitute for it.

5. Escalate continued impairment as an exposure, not an overdue task

Leaders should not only see how many critical-control actions are late. They should see:

  • which critical controls are currently impaired;
  • what hazardous work has stopped;
  • where temporary arrangements are being relied upon;
  • who approved those arrangements;
  • how long the exposure has existed;
  • what verification is required before restart; and
  • whether similar control weaknesses could exist elsewhere.

That is materially different from a maintenance backlog.

Stop-work authority is not enough

Many organisations say every worker has the authority to stop unsafe work. That is important. But authority on paper does not guarantee that stopping is practical.

Workers will interpret what the organisation really expects from:

  • how supervisors respond when work is paused;
  • whether production targets remain unchanged;
  • whether people are blamed for “overreacting”;
  • how contractors are treated when they raise defects;
  • whether temporary workarounds become normal; and
  • how quickly leaders remove pressure from the person who made the call.

The better question is not simply, “Are workers allowed to stop?”

What in our system makes stopping the normal, supported response when a critical control is unavailable?

What officers should ask to see

South Australia’s industrial manslaughter offence applies to PCBUs and officers and carries a maximum penalty of 20 years’ imprisonment for an individual and $18 million for a body corporate. SafeWork SA emphasises that the offence did not create new WHS duties; it attaches the most serious consequences to particular breaches of existing duties that cause death and involve gross negligence or recklessness.[5][6]

The lesson for governance is not to wait for a prosecution and then examine one event in isolation. Officers should ask:

  • Do we know which controls prevent our highest-consequence events?
  • What automatically happens when one is reported as impaired?
  • Can production continue while a critical-control work order remains open?
  • Who can authorise an alternative arrangement?
  • What evidence is required before the plant or task returns to service?
  • Can we see current impairment and exposure, not only overdue maintenance?
  • Have we tested whether the process works on nights, weekends and under production pressure?

These questions turn due diligence from a review of completed paperwork into assurance about present operating conditions.

The decision that must not disappear

A technician can identify a problem. A worker can report it. A maintenance system can record it. A dashboard can display it. None of those actions, by themselves, controls the exposure. The control is effective only when the information changes what the organisation does.

If a critical control failed today, who would stop the work, and what prevents anyone from quietly starting it again?

This article provides general information, not legal advice. The South Australian proceedings discussed are active. The allegations have not been tested, and no inference about guilt or responsibility should be drawn from this commentary.

References

  1. Liam O’Connor, LinkedIn post: South Australian manufacturer and directors charged, accessed 29 July 2026. The post contains the supplied 7NEWS footage and a summary of allegations; it is used for media context, not as a judicial finding.
  2. ABC News, “Industrial manslaughter charges laid against Adelaide company after 2024 workplace death”, 27 July 2026. Reports the filed charges and allegations contained in court documents. The matter remains before the courts.
  3. SafeWork SA, “Guarding”, accessed 29 July 2026.
  4. SafeWork SA, “Guard against unsafe machinery”, June 2026.
  5. SafeWork SA, “Industrial manslaughter”, accessed 29 July 2026.
  6. South Australian Legislation, Work Health and Safety Act 2012, current authorised version, accessed 29 July 2026.
  7. SafeWork SA, Managing the risks of plant in the workplace: Code of Practice, accessed 29 July 2026.