Danny has spent fourteen years on the potlines at an aluminium smelter on the New South Wales coast, the kind of job that pays well enough to raise a family in a town where the smelter is one of the largest employers for fifty kilometres in any direction. He has watched successive governments promise to fix the industry's energy costs, and in December 2025 he watched the Commonwealth agree to underwrite his employer's power bill for the next decade, at a cost that could run into the billions, to keep the site open (approx.; based on public reporting of the Tomago Aluminium government-backed power agreement, announced Dec-2025). Danny is not a climate sceptic and he is not indifferent to emissions — he just wants to know whether the rules his employer now operates under are actually going to keep people like him in work, or whether they are a slow, well-intentioned way of moving his job, and the pollution that comes with it, to a smelter in a country that does not ask the question at all.
Danny's employer is one of roughly 208 large industrial sites across Australia now covered by the Safeguard Mechanism — the scheme that sets a legal, gradually shrinking limit on how much carbon dioxide the country's biggest individual polluters can emit (Clean Energy Regulator, 2024-25 Safeguard Mechanism data, published 2026). Steel mills, aluminium smelters, cement kilns, LNG plants and coal mines make that list, and together they are responsible for well over a quarter of Australia's total greenhouse gas emissions (approx.; Clean Energy Regulator, 2023-24 Safeguard Mechanism data). Since July 2023, each facility has operated under a baseline — an emissions ceiling specific to that site — that the government tightens by law, year after year, on a trajectory that gets steeper toward 2030. For most businesses on the list, that decline is real and largely uncushioned. For a smaller group — steelmakers, aluminium smelters and cement producers competing directly against factories overseas that pay no carbon cost at all — the government has built in slower declines and direct taxpayer support, on the argument that pushing them too hard, too fast, will not cut a single tonne of global emissions. It will just close the plant and let the same production, and the same or dirtier emissions, continue somewhere with weaker rules. That argument is either the reformed Safeguard Mechanism's most defensible design choice, or its biggest loophole. This is the evidence for judging which.
What the government did
The Safeguard Mechanism did not start with the Albanese government. It began under the Abbott government's Direct Action climate policy, commencing on 1 July 2016 as the compliance backstop sitting alongside the Emissions Reduction Fund, made under the National Greenhouse and Energy Reporting Act 2007 (Cth). Its design was simple on paper: any facility emitting more than 100,000 tonnes of carbon dioxide-equivalent (CO2-e) a year — a large multiple of a typical Australian household's or vehicle's annual footprint (approx.) — had to keep its emissions under a baseline set for that site (National Greenhouse and Energy Reporting Act 2007 (Cth) and the Safeguard Mechanism rules made under it). In practice, those original baselines were set at or near each facility's highest year of emissions over the prior several years, with provisions to reset them higher again if a business grew. Climate policy analysts and parts of the energy industry itself described the result bluntly: in the scheme's first six years, breaches were rare and penalties rarer still, because the ceiling sat close to or above what a facility was already emitting (approx.; secondary analysis by energy-sector commentators and climate policy researchers, 2016–2022). It was, on this widely shared reading, a safeguard against very little.
Labor took a commitment to reform the Safeguard Mechanism to the 2022 election as part of its Powering Australia plan, and it legislated the destination first: the Climate Change Act 2022, passed in September 2022, wrote Australia's targets of a 43 per cent cut in emissions below 2005 levels by 2030 and net zero by 2050 into law (Climate Change Act 2022 (Cth)). The Safeguard Mechanism reform was the vehicle chosen to make roughly 215 to 219 of the country's biggest individual industrial emitters do their share of that task, rather than leaving it entirely to the electricity grid's shift to renewables (Department of Climate Change, Energy, the Environment and Water (DCCEEW), Safeguard Mechanism reforms factsheet, 2023). The reform passed Parliament in March 2023 as the Safeguard Mechanism (Crediting) Amendment Act 2023, after the government negotiated amendments with the Australian Greens to secure the Senate crossbench votes it needed, and took effect on 1 July 2023.
The mechanics are a baseline-and-credit system — a lighter-touch cousin of a full emissions trading scheme. Each covered facility gets an annual baseline, its personal emissions ceiling, which now declines automatically by a default 4.9 per cent every year through to 2030, a rate set to keep the covered facilities collectively on track for the economy-wide 43 per cent target (Safeguard Mechanism rules, as amended 2023; DCCEEW, Safeguard Mechanism overview). A facility that emits under its declining baseline earns tradeable Safeguard Mechanism Credits (SMCs) it can sell to others; a facility that emits over its baseline must acquit the gap by surrendering SMCs it has bought, or Australian Carbon Credit Units (ACCUs) — tradeable certificates, one per tonne of CO2-e, generated mainly by activities such as regrowing native vegetation, capturing landfill gas and changing land management, and issued by the Clean Energy Regulator under the Carbon Credits (Carbon Farming Initiative) Act 2011. To stop compliance costs spiralling if ACCU prices spike, the reform also built in a cost containment measure: a liable facility can instead buy an ACCU directly from the Clean Energy Regulator at a fixed price, set at $75 a tonne in 2023-24 and indexed each year to inflation plus 2 per cent, reaching $82.68 a tonne in 2025-26 (Clean Energy Regulator, cost containment measure; DCCEEW, Safeguard Mechanism reforms factsheet, 2023) — in effect, a price ceiling on how much the scheme can cost any single business in any single year.
The Greens' price for their Senate votes went further than the decline rate. The final legislation includes a legislated absolute ceiling — a "hard cap" — on the combined emissions of every facility in the scheme: no more than 1,233 million tonnes of CO2-e cumulatively through to 30 June 2030, tightening to no more than 100 million tonnes in the 2029-30 financial year alone, which the Greens and government both described as roughly 28 per cent of the entire economy's remaining carbon budget over the same period (Safeguard Mechanism (Crediting) Amendment Act 2023; Australian Greens, media statements, Mar-2023). Any new facility, or any expansion of an existing one — including new coal and gas projects — must have its additional emissions fully offset so the hard cap is not breached, giving the government a lever, for the first time under this scheme, to constrain new fossil fuel developments on climate grounds alone (approx.; Australian Greens, media statements, Mar-2023).
For the subset of covered facilities the government judges to be trade-exposed — competing against overseas producers who face no equivalent carbon cost — the scheme builds in a second, gentler track. A facility can apply to the Clean Energy Regulator for Trade-Exposed Baseline-Adjusted (TEBA) status once the estimated cost of compliance exceeds 3 per cent of its earnings before interest and tax (EBIT); once approved, its baseline can decline at a reduced rate — as low as 1 per cent a year for manufacturing facilities such as steelworks, aluminium smelters and cement plants, or 2 per cent a year for other trade-exposed sectors, instead of the standard 4.9 per cent — for renewable three-year terms, with the lowest rates reserved for facilities where the compliance cost reaches 8 to 10 per cent of EBIT, depending on the sector (Safeguard Mechanism rules, TEBA provisions, 2023; Clean Energy Regulator, TEBA facility data). Alongside that slower baseline path, the government set up the $600 million Powering the Regions Fund — Safeguard Transformation Stream, competitive grants of between $500,000 and $50 million to help exactly these trade-exposed facilities pay for on-site decarbonisation — new kilns, electrified furnaces, process redesigns — rather than simply buying their way to compliance with credits (DCCEEW, Powering the Regions Fund — Safeguard Transformation Stream, program guidelines).
The strongest case for the policy
Put the government's defence at its best
The mechanism now doing the work is a Coalition-designed instrument, not a new Labor invention, and that lineage matters practically as well as politically. Australia spent 2011 to 2014 tearing itself apart over an economy-wide carbon price, introduced by one government and repealed by the next, and businesses making twenty- and thirty-year investment decisions have said clearly, and repeatedly, in Safeguard Mechanism submissions and reviews, that policy stability matters to them as much as any single design detail (approx.; industry submissions to the Safeguard Mechanism reforms consultation, 2022-23). By reforming an existing, already-legislated scheme that industry had lived under since 2016, rather than starting again from scratch, the government reduced the risk of another decade-long fight over the fundamental existence of a carbon constraint. On the substance, the reform did what its predecessor never did: it converted a ceiling almost no facility ever hit into one that declines automatically, by law, every year, for the biggest individual industrial polluters in the country — the source of more than a quarter of the nation's emissions. And it did not pretend a steel mill and a data centre are the same kind of business. Facing government-verified evidence that a facility competes against producers in countries with no carbon cost, the scheme slows the decline rate, caps the compliance cost, and puts real money on the table to help that facility cut emissions on-site — because a government that ignored the trade-exposed problem would either see those plants close, with the same production and the same emissions continuing somewhere with weaker rules, or would see the policy quietly abandoned the moment a marginal seat's steelworks was threatened. Tailored treatment for trade-exposed industry is not a carve-out from climate ambition; on this reading, it is the design feature that lets the ambition survive contact with the real economy.
The evidence: what the primary record shows
The scheme, in the regulator's own numbers
The Clean Energy Regulator publishes Safeguard Mechanism compliance data every year, and the first two years of the reformed scheme give an early, honest read on how it is working. In 2023-24, the first year of the new declining baselines, 219 facilities were covered, with combined reported emissions of 136.0 million tonnes of CO2-e (Clean Energy Regulator, 2023-24 Safeguard Mechanism data, published Apr-2025). In 2024-25, the number of covered facilities fell to 208 and their combined reported emissions fell to 132.8 million tonnes of CO2-e, a drop of about 2.3 per cent (Clean Energy Regulator, 2024-25 Safeguard Mechanism data, published Apr-2026). Compliance was high in both years — 205 of the 208 covered facilities, or 98.6 per cent, were not in an excess-emissions position at the 2024-25 deadline. On the credit side, 54 facilities earned a combined 6.7 million Safeguard Mechanism Credits in 2024-25 for beating their baseline, down from 64 facilities and 8.3 million credits the year before (Clean Energy Regulator, 2024-25 Safeguard Mechanism data).
Read one way, that is the scheme working as designed: total covered emissions falling, in law, for the first time since the mechanism existed. Read more carefully, the numbers also show a gap between the legislated decline rate and what is actually happening in facility smokestacks. A 4.9 per cent default decline rate should, on its own, produce a fall of roughly that size in the underlying baseline pool each year. Reported covered emissions fell by about 2.3 per cent in 2024-25 — under half that rate. Some of the difference is legitimate: not every facility sits on the default rate, part of "compliance" comes from surrendering credits or ACCUs rather than cutting emissions at the source, and the scheme lets facilities bank and carry credits between years. But the gap is also the first hard evidence for a question this topic exists to test: whether a legislated decline rate on paper is the same thing as an equivalent fall in real industrial emissions.
The ACCU market behind the offset half of the scheme
Every tonne a facility emits above its baseline has to be covered by a credit, and for most facilities that credit is an Australian Carbon Credit Unit, not a Safeguard Mechanism Credit bought from another industrial site. ACCUs are generated mostly by land-based projects — regrowing native vegetation, capturing landfill gas, changing how farmland or forests are managed — registered and issued by the Clean Energy Regulator under the Carbon Credits (Carbon Farming Initiative) Act 2011. Their price has moved substantially since the reform began: the spot price traded in a band of roughly $32 to $38.50 a tonne for most of 2024, spiked to about $42.55 in mid-November 2024, then eased back to trade around $35 to $38 a tonne through most of 2025 (approx.; Clean Energy Regulator, quarterly carbon market reports, 2024–2025). That price stayed well below the government's $75-to-$82.68 cost containment ceiling through the reform's first two years, meaning the fixed-price safety valve has not yet had to be used by most liable facilities — the open market has been cheap enough to do the job (approx.; Clean Energy Regulator, quarterly carbon market reports). The Clean Energy Regulator issued a record 18.8 million ACCUs in 2024, up 9 per cent on the 17.2 million issued in 2023, with first-quarter 2025 issuance of 3.0 million units, putting the scheme on track for an estimated 19 to 24 million ACCUs across 2025 (Clean Energy Regulator, ACCU Scheme quarterly issuance data, 2023–2025).
Whether those tonnes represent genuine, additional abatement — carbon that would not have been stored or avoided without the scheme — has been formally tested once already. In June 2022, then-Minister for Climate Change and Energy Chris Bowen appointed former Chief Scientist Professor Ian Chubb to lead an independent review of the ACCU scheme's integrity, reporting in December 2022. The Chubb Review found the scheme's basic architecture "essentially sound," rejecting claims that abatement had been systematically overstated, but it also commissioned the Australian Academy of Science to independently examine four specific methods — human-induced regeneration of native forest, avoided deforestation, landfill gas capture, and carbon capture and storage — precisely because their additionality, the question of whether the claimed carbon storage would have happened anyway, had been challenged by scientists and journalists (Independent Review of ACCUs, "the Chubb Review," final report, Dec-2022, DCCEEW). The review recommended governance and transparency reforms, most of which the government has since moved to implement, but the underlying tension it examined — that a scheme built substantially on land-sector credits is harder to audit at the tonne level than a scheme built on metered smokestack reductions — has not disappeared. It matters directly here because trade-exposed facilities on slow, TEBA-adjusted decline rates are, by the scheme's own design, more likely to lean on bought credits rather than on-site process change to stay compliant in the near term, which means the credibility of their contribution to the declining national total rests partly on the credibility of a market that has already needed one independent integrity review.
A legislated 4.9 per cent decline rate and an actual 2.3 per cent fall in reported emissions are not the same fact, and the gap between them is where this topic lives.
What "trade-exposed" looks like on the ground: three case studies
The abstract category of "emissions-intensive trade-exposed industry" has a very concrete face in three separate government interventions since the reformed scheme began. In February 2025, the Whyalla steelworks in South Australia — owned by Sanjeev Gupta's GFG Alliance and one of only two primary steel producers left in the country — was placed into administration after the South Australian Parliament amended the Whyalla Steel Works Act 1958 (SA) to force the move, with creditors owed more than $1 billion, including roughly $500 million to GFG entities themselves (KordaMentha, administrator's reports; South Australian Parliament, amendments to the Whyalla Steel Works Act 1958, Feb-2025). The Commonwealth and South Australian governments jointly announced a $2.4 billion support package the following day, 20 February 2025, to keep the plant running while a buyer is found (Australian Government and South Australian Government, joint media statement, 20-Feb-2025). Whyalla's collapse had company-specific roots — GFG Alliance's broader debts and its owner's long-running financing troubles predate the Safeguard Mechanism reform by years — and this topic does not claim the reformed scheme caused it. But it is a real, current illustration of how thin the margin already is at Australia's oldest heavy-industry sites, and of a government stepping in with billions of taxpayer dollars specifically to keep a trade-exposed, emissions-intensive facility from closing.
Rio Tinto's Boyne Smelters at Gladstone, Queensland — a 500,000-tonne-a-year aluminium smelter — received a similar signal in 2026, when the Commonwealth and Queensland governments each committed roughly $1 billion over ten years, a combined $2 billion (approx.), to support the plant's shift to renewable power ahead of its existing electricity contracts expiring in 2029 (approx.; Australian Government and Queensland Government, media statements, 2026). Rio Tinto has separately contracted 2.7 gigawatts of renewable generation and 540 megawatts of battery storage, projected to supply about 80 per cent of the smelter's average electricity demand and cut its direct and electricity emissions by around 70 per cent once complete (approx.; Rio Tinto, market disclosures, 2025–2026). Tomago Aluminium on the New South Wales Hunter coast — Australia's largest single electricity user, drawing roughly 10 per cent of the state's total electricity supply and producing more than a third of the country's aluminium — is Danny's employer, and it received its own government-backed power deal in December 2025, under which the Commonwealth-owned Snowy Hydro will buy renewable electricity at market rates and on-sell it to the smelter at a subsidised price, with total fiscal exposure over the ten-year deal estimated at $3 billion to $4.7 billion, or roughly $300 million to $470 million a year (approx.; Australian Government, Tomago power agreement announcement, Dec-2025; industry and government estimates reported at the time).
Three of Australia's most significant trade-exposed industrial sites, in three different states, all received direct government financial intervention within about twelve months of each other. None of that support flows through the Safeguard Mechanism's own mechanics — it sits in energy policy and industry assistance, not the baseline-and-credit system itself — but it cannot be read in isolation from it. The government is not simply setting a slower emissions-decline rate for these industries and letting the market sort out the rest. It is also putting billions of taxpayer dollars directly into keeping the same handful of plants open, which raises a fair question about which lever is actually preventing closure and job losses: the softened Safeguard Mechanism treatment, or the parallel industry-support cheques.
The government's own carbon leakage review
The government did not leave the trade-exposed question to guesswork. Between July 2023 and February 2025, an independent Carbon Leakage Review led by Australian National University economist Professor Frank Jotzo assessed the leakage risk facing all 75 trade-exposed commodities across 42 commodity groups covered by the Safeguard Mechanism, out to 2030 (Carbon Leakage Review, final report, DCCEEW, 13-Feb-2025). Its headline finding cuts both ways for this topic's central question. The review concluded that Australia's existing settings — the slower TEBA decline rates and the transformation funding — are effective at containing leakage risk for most trade-exposed industry in the short-to-medium term. But it identified a smaller group of commodities where the risk is real and material: cement and clinker were assessed as most exposed, with a potential rise in imports of around 14 per cent by 2030 under the review's maximum-risk scenario, followed by lime; a second tier — steel and iron, ammonia and derivatives, glass and hydrogen — also carries genuine leakage risk, though a more complex one to design a fix for, given varied production methods and supply chains (Carbon Leakage Review, final report, DCCEEW, Feb-2025). The review's central recommendation was not to weaken the Safeguard Mechanism's ambition for those sectors, but to introduce a border carbon adjustment — a tariff-like charge on imports of these goods from countries with weaker climate policies, similar in concept to the European Union's scheme — staged, starting with cement and clinker before expanding to the others. None of that recommendation has yet been legislated, and the government has flagged that the suitability of the current TEBA arrangements for these specific commodities will be reconsidered as part of a wider Safeguard Mechanism review beginning in the second half of 2026, with a consultation paper expected around July 2026 (DCCEEW, 2026-27 Review of the Safeguard Mechanism, consultation timeline). That review will also have to decide the baseline decline trajectory beyond 2030 — a decision not yet made in law, which matters enormously to any business weighing a multi-decade investment in a steel mill, smelter or cement kiln today.
The wider world is starting to price this too
Australia's leakage debate is not happening in isolation. On 1 January 2026, the European Union's Carbon Border Adjustment Mechanism (CBAM) entered its definitive phase, requiring EU importers of cement, iron and steel, aluminium, fertilisers, hydrogen and electricity to buy certificates covering the embedded carbon in those goods, priced in line with the EU's own carbon market — the first quarterly certificate price was set on 7 April 2026 at €75.36 a tonne of CO2-equivalent (European Commission, Carbon Border Adjustment Mechanism, 2026). Analysts assess Australia's direct trade exposure to the EU mechanism as limited, because relatively little Australian steel, aluminium or cement is sold directly into the European market (approx.; trade and carbon-market analysis, 2025–2026). But the direction of travel matters more than any single country's exposure: as more trading partners begin pricing embedded carbon at the border, the case for indefinitely protecting Australian trade-exposed industry from any domestic carbon cost gets weaker over time, even as the case for getting the domestic-versus-imported cost gap right in the meantime gets stronger.
The critique: two ways to lose the same argument
The steel-man case for tailored trade-exposed treatment is genuine, and the evidence above does not overturn it: a Safeguard Mechanism that ignored the competitiveness problem would very likely see Whyalla, Boyne and Tomago close years sooner, with the same steel and aluminium produced somewhere with a dirtier grid and no carbon constraint at all — a worse outcome for the planet, not a better one. But the same evidence exposes two separate ways the current design can fail on its own terms, and both point toward the same underlying risk: an Australia that exports its emissions and its jobs at once, while the headline Safeguard Mechanism numbers still show a national win.
The first failure mode is arithmetic. A facility on the minimum TEBA decline rate of 1 per cent a year is, by construction, barely declining at all against a scheme built around a 4.9 per cent default. Over the six years to 2030, a 4.9 per cent annual decline compounds to roughly a 26 per cent fall in baseline; a 1 per cent annual decline compounds to less than 6 per cent. That gap is not a design accident — it is the explicit, government-approved trade-off for facilities judged genuinely at risk of carbon leakage. But it means the national decline figure the government can point to is carried disproportionately by facilities that are not trade-exposed, while the sectors that produce some of the most emissions-intensive goods in the economy — steel, aluminium, cement — do comparatively little of the heavy lifting inside the scheme itself. The Carbon Leakage Review's own finding that cement and clinker face a genuine, material leakage risk, even under current protective settings, is the government's own evidence that the trade-exposed track is not yet a solved problem — it is a live one, still under review, with the answer due sometime after mid-2026.
The second failure mode is where the credit sits. For a facility on a slow decline rate that still exceeds even that reduced baseline, compliance runs through the ACCU market rather than through cutting emissions at the smokestack. That is by design — a baseline-and-credit scheme is meant to let the cheapest abatement happen first, wherever it is — but it means a trade-exposed facility can report full Safeguard Mechanism compliance in a given year without changing its own production process at all, so long as it has bought enough tonnes of vegetation regrowth or landfill gas capture from somewhere else in the country. The Chubb Review found the ACCU scheme's overall architecture sound, but it did not find every method beyond question, and it specifically flagged additionality concerns in exactly the land-sector methods that supply most of the market's volume. A national number built partly on offsets whose integrity has already needed one independent review is not the same as a national number built on measured reductions at 208 individual smokestacks, even when both numbers say "emissions are falling."
A steelworks that buys its way to compliance and a steelworks that re-engineers its furnace both show up as the same green tick in the annual data. Only one of them has actually changed what comes out of the stack.
Put those two failure modes together with the direct government interventions at Whyalla, Boyne and Tomago, and a harder version of the leakage question comes into view. The Safeguard Mechanism's trade-exposed provisions are meant to prevent carbon leakage by asking less of these facilities than the standard scheme would. In parallel, separate energy and industry-support programs are pouring billions of taxpayer dollars into the same handful of sites to prevent them closing anyway. If both the softened Safeguard Mechanism treatment and billions of dollars in direct subsidy are needed just to keep Australia's aluminium and steel capacity standing still, that is a signal the underlying economics of running these plants in Australia — largely a question of electricity price and grid reliability, only partly a question of the Safeguard Mechanism itself — were already fragile before the reformed scheme arrived. The risk is that the trade-exposed carve-out gets credited, politically, with saving jobs and industry that a much larger and more expensive set of energy-policy interventions is actually doing the work of saving, while the climate policy itself quietly asks very little of the country's largest point-source industrial emitters.
What they got right, and the honest caveats
Credit is due on several specific points. Converting a scheme that almost never bound any facility into one with a legislated, compounding, publicly reported annual decline is a genuine structural improvement, not a cosmetic one — the 2024-25 data, whatever its imperfections, shows covered emissions falling for the first time since the mechanism existed, and 98.6 per cent of facilities were compliant with a real, tightening rule rather than a notional one (Clean Energy Regulator, 2024-25 Safeguard Mechanism data). The hard cap the Greens negotiated is a genuinely binding constraint that did not exist before 2023: an absolute cumulative ceiling of 1,233 million tonnes to 2030, with new projects required to fully offset their additional emissions, is a real check on new coal and gas developments the original scheme never provided. The cost containment measure is sound, orthodox economic design — it protects businesses, and by extension jobs, from an uncontrolled spike in compliance costs without removing the underlying obligation to reduce emissions over time. And commissioning the Carbon Leakage Review before finalising how to treat the highest-risk trade-exposed commodities, rather than guessing, is the kind of evidence-based policymaking this site credits when it happens: the review identified specific at-risk sectors — cement, steel and a handful of others — rather than treating all "heavy industry" as equally exposed, and it proposed a specific fix, a border carbon adjustment, rather than simply softening the domestic rules further.
The honest caveats sit mostly in what has not yet been decided. The baseline decline trajectory beyond 2030 is not yet legislated, which is a real problem for businesses planning investments — a new furnace or kiln — on a twenty- or thirty-year horizon; the 2026-27 Safeguard Mechanism review, only now getting under way, will have to answer that question, along with whether the TEBA settings for cement, steel and the other flagged commodities need tightening. The Carbon Leakage Review's recommended border carbon adjustment remains just that — a recommendation, not law — and its cement-first, staged rollout means the highest-risk commodity might be the only one actually protected by a border measure for some years yet. And it is fair to note, as this site's analysis of the 43% target and the net-zero arithmetic and the real cost of the transition sets out in more detail, that much of the pressure squeezing Tomago, Boyne and Whyalla flows from electricity prices and grid reliability during the broader energy transition, not from the Safeguard Mechanism specifically — which means fixing the Safeguard Mechanism's trade-exposed settings alone will not, on its own, resolve the deeper competitiveness problem heavy industry is living through.
The risk to the future
A cleaner ledger, not necessarily a cleaner planet
The risk this topic is built around is not that the Safeguard Mechanism fails loudly. It is that it succeeds quietly, on paper, while the underlying problem it exists to solve — global emissions from steel, aluminium and cement production — simply relocates. If the trade-exposed settings stay too generous for too long, or if compliance keeps leaning on an offset market whose integrity has already needed one independent review, Australia can report a declining national industrial emissions total even as individual plants close, or quietly shrink production, and the same tonnes of steel and aluminium get made in a country with a dirtier grid and no equivalent scheme at all. That is carbon leakage in its purest form: a domestic win on the spreadsheet, a wash or a loss for the atmosphere, and a real loss of jobs and industrial capability in towns — Whyalla, Gladstone, the Hunter — that do not have an easy second industry to fall back on. The 2026-27 Safeguard Mechanism review, and the still-unlegislated decisions on a border carbon adjustment and the post-2030 decline trajectory, will do more than any other single policy choice to determine which future Australia gets: a heavy industry sector that genuinely re-engineers itself onto cleaner processes and keeps the jobs, or one that is quietly wound down, offshored, and replaced by imports the country then has no way to green at all. The next generation inherits either a smaller, cleaner industrial base, or a hollowed-out one sitting behind a rising import bill — and on the evidence collected here, which outcome Australia gets remains genuinely open.