There is a particular moment on any site where work happens above the ground. A supervisor stands at the base of a scaffold, tilts their head back, and takes in the whole arrangement at once — the guardrail that stops a metre short of the corner, the anchor point bolted into something that has not yet been braced, the worker on the roof edge whose lanyard is clipped to a purlin someone assumed would hold. Most of the time, nothing happens. The job finishes, the crew goes home, and the arrangement that was almost right stays almost right.
For years, the law's response to that moment carried a certain softness. Codes of practice explained how falls should be managed, and a business that ignored them was left exposed to an argument about whether it had done what was reasonably practicable. That argument is now much shorter.
On 1 July 2026, section 26A of the *Work Health and Safety Act 2011* (NSW) came into effect, inserted by Schedule 3 of the Industrial Relations and Other Legislation Amendment (Workplace Protections) Act 2025. Its wording is plain. A business must comply with an approved code of practice — or manage the same hazards in a different way that delivers a standard of health and safety equivalent to or higher than the code requires.
Read that second limb carefully, because it is where the real shift sits. You may still depart from a code. What you may no longer do is depart from it by accident, or by habit, and explain it afterwards. The burden of showing that your alternative is genuinely equal or better now sits with you, in advance, in writing.
Approved codes have gone from being the thing a regulator quotes to being the thing a regulator measures you against.
Eight NSW codes of practice carry a May 2026 date, and one of them, gazetted and in force from 22 May, is *Work on roofs – commercial and industrial buildings*. It replaces the 2009 *Safe Work on Roofs* code, and anyone who works above ground level should read it — not because it is surprising, but because it is specific in the places where site practice tends to be vague.
It restates the hierarchy of control without apology: eliminate the need to work at height, then prevent the fall with edge protection, then use work positioning, and only as a last resort arrest the fall. Guardrails must be installed by competent people, and the code draws attention to the geometry — the gaps between rails, the distance from the edge — that quietly opens up as work progresses and a system installed correctly on Monday stops being correct by Thursday.
On anchor points it is blunter still. Anchorages must be verified by a competent person as capable of taking fall-arrest loads. A roof truss counts only if its capacity has been confirmed. Structures that are temporarily braced or incomplete are not anchor points, however convenient they look.
And then there is the line that ought to change how some sites are staffed. The code's position is that a worker should not use a fall arrest system unless at least one other person on site is capable of rescuing them. A harness is not a control measure on its own. A harness plus a plan plus somebody able to execute it is a control measure. Suspended in a harness, a worker's own survival time is measured in minutes, not the half hour it takes for an ambulance to find the gate, get through the induction and reach the base of the structure.
SafeWork NSW named its four regulatory priorities for 2026-27 in July, and falls from heights sits at the top, alongside psychosocial risks, exposure to hazardous substances such as asbestos, crystalline silica and welding fumes, and injury from mobile plant, vehicles and fixed machinery. The reasoning is not abstract. More than 600 falls-from-height incidents were reported in the preceding twelve months. Five of them killed someone.
More uncomfortable is SafeWork's own evaluation of its 2025 falls from heights program, which visited 935 workplaces — most of them construction sites. Inspectors issued 1,058 compliance notices and $223,950 in on-the-spot fines. Forty-five per cent of the businesses visited did not have an adequate system in place for managing the risk of falls. Scaffolding was the single worst-performing control, non-compliant at nearly half of the sites where it was inspected. Other building edges were close behind at 47 per cent, and roofs themselves at 40.
The same report contains a detail worth sitting with. Eighty-two per cent of those businesses were found to have given workers adequate information, instruction and training, and the same proportion had consulted them properly. Seventy-nine per cent were supervising work at heights appropriately. In other words, the training was largely in place and the systems around it were not. Competence and control are related, but they are not the same thing, and a workforce that knows exactly what good looks like can still be sent up a scaffold that is wrong.
The practical test of section 26A is whether you can produce an answer, not an intention. For work above ground, three questions carry most of the weight:
Can you show the control, not just the policy?
A safe work method statement that names edge protection is not evidence that edge protection existed at the edge that mattered, on the day, after the last stage of work changed the geometry.
Can you show the equipment is fit to catch someone?
Harnesses, lanyards, inertia reels, static lines and anchor points all have inspection regimes and service lives. AS/NZS 1891.4 sets the expectation — the roof code points to the 2025 edition — and six-monthly inspection by a competent person is the practical floor. Anchor points and static lines need testing and certification, and the certificate is the part that survives being asked about.
Can you show someone could get them down?
Not call for help — get them down. If the answer depends on a person who is rostered elsewhere, or on a rescue procedure that has never been run, then the fall arrest system is doing less work than the paperwork suggests.
Preparation, not paperwork
None of this is a reason for alarm, and it is not a new burden invented by a regulator looking for revenue. Most of what the updated roof code asks for is what careful operators were already doing: competent training, verified equipment, a plan that has been tested rather than filed.
What has changed is the evidentiary position. The code is the standard now. The question an inspector asks after a fall — and increasingly, before one — is not whether you meant well, but what you can show.
Zokal's working at heights training delivers RIIWHS204E on simulation towers and includes emergency response and height rescue, with a refresher for crews whose competency has aged out. Our technicians handle height safety equipment inspection, servicing and anchor point certification to AS/NZS 1891.4, and issue the service reports and compliance certificates that make the second question above answerable. And where a site needs rescue capability standing by rather than written down, our height safety standby personnel are trained in vertical rescue and casualty retrieval, and can either run the rescue or support your own emergency response team.
The supervisor at the base of the scaffold, looking up: that instinct was always the right one. The law has simply caught up with it.