Warning Time — 18 September 2026

AUKUS turned five this week, and the commentariat marked the anniversary the way Australians mark most things: by arguing about whether we should have bothered. Paul Keating says the deal locks us into a war we'd "certainly lose". An ambassador broke ranks. Meanwhile the UK is about to cut steel at Barrow and the PLA rammed a Philippine government vessel. The debate is loud; the clock is louder.

The Big Movements

AUKUS at five: the politics wobble, the steel doesn't

Three things happened at once. Keating went public arguing AUKUS commits Australia to a losing American war against China. A serving-or-former ambassador broke ranks on China and AUKUS policy. And Dennis Richardson's successors at the top of the defence establishment — Moriarty among them — went out swinging to say the programme is on track. Across the Atlantic, the UK is set to cut first steel on its AUKUS boats at Barrow next year, and a former White House official conceded the thing has felt "too hard" at times but survived anyway.

The fault line isn't political will. It's industrial capacity and workforce, in all three countries, five years in and still unresolved. That's the finding that matters more than any op-ed. You cannot argue a submarine into existence, and you cannot recruit a welder with a press release. Keating's critics and Keating's defenders are having a debate about strategy when the binding constraint is people and factories.

Judge AUKUS by build rate, trained tradespeople and Henderson throughput — not by who won this week's column inches. If the workforce numbers don't move in the next two budgets, the sceptics will be proved right for reasons that have nothing to do with their argument.

China is industrialising coercion at sea — and Washington is looking elsewhere

A Chinese vessel rammed a Philippine government ship in the South China Sea. Separately, analysis this week describes something more disciplined: an institutionalised Chinese approach to maritime coercion around Taiwan, normalising PLA and Coast Guard presence and establishing effective control of shipping lanes without firing a shot. And The Australian reports Xi is positioning to exploit a US president perceived as cooling on Taiwan.

Put those three together and you get the actual threat model. Not an invasion on a Tuesday. A slow ratchet that tests allied tolerance, one incident at a time, while the guarantor of the system signals it may not turn up. Australia formally pressed Beijing over its submarine-launched ballistic missile test during defence talks this week and warned of escalation risk. Good. It changes nothing on the water.

Australia's response can't be another communiqué. It has to be presence — persistent, unglamorous, in company with Japan, the Philippines and Korea — plus the long-range strike and sea denial capability that makes a fait accompli expensive. Deterrence is a budget line, not a sentiment.

They're already inside the grid

The ABC set out this week how China could hack Australia's power and telecommunications networks today. Not in a future contingency. Now. Pre-positioned access in the systems that run electricity and comms is the standard opening move in any serious conflict scenario, and it is a capability Australia has no kinetic answer to.

Alongside it: China is building an AI system to tailor foreign propaganda, calibrated using data from Australian respondents. That's the shift from broadcast propaganda to targeted cognitive warfare, aimed squarely at our political cleavages. Note the grim symmetry — the same week ASD's director called for Australia to train AI models at home rather than rent capability from someone else's ecosystem.

Two prescriptions, both boring, both overdue. Mandate and audit critical-infrastructure hardening with actual consequences for failure, because voluntary frameworks have not worked. And back ASD's call: sovereign AI training capacity is not industry policy, it's a defence requirement. Australia cannot contest an information war using foundation models it doesn't control.

The fiscal engine is losing compression

ASPI put a number on it this week: Australian iron ore export earnings are on track for a 34 per cent real-terms decline by 2030–31, down to A$77.2 billion. Volumes barely move. It's price compression, and it reflects China's improving leverage as a buyer.

This is the story nobody wants to lead with because it isn't a warship. But every dollar of defence expansion, every submarine, every missile factory, is underwritten by a resources sector whose single biggest customer is also our principal strategic competitor. That's not a supply chain. It's a hand around the throat of the budget.

Two responses. Diversify who buys the ore and what we sell them — processed product, not just dirt on a boat. And stop pretending defence spending decisions can be made independently of commodity price assumptions. If the revenue forecast is wrong, the force structure is wrong.

Canberra finally builds the thing it should have built a decade ago

Labor is establishing a dedicated defence-manufacturing agency — the AFR framed it as building Australia's own Lockheed Martin. In the same week: the ADF workforce hit its largest size in 25 years, and there's serious talk of an Australian defence export boom.

Three genuine positives, and I'll take them. Sovereign manufacturing capacity is the pillar where Australia has been weakest for longest, and it's the constraint that's throttling AUKUS delivery. A workforce high is meaningful because personnel, not platforms, is the harder number to fix.

The test is whether the new agency gets the authority to break acquisition orthodoxy or becomes another coordinating body writing strategies about coordination. ASPI's argument this week — look to commercial providers for fast capability — is the right instinct. Buy what exists, field it now, fix it later. The perfect acquisition delivered in 2040 is worth nothing.

Also on the Watch

• Senate demands an outside review of the inquiry into the fatal MRH-90 Taipan crash — trust in ADF accident investigation is now itself a readiness problem. The Australian
• ASD's director says Australia must train AI at home to protect national security. AFR
• AUKUS Pillar II — quantum, cyber and AI — is the part of the deal that can deliver before 2040. AFR
• US trade chief pressed Brazil to review Anglo American's nickel sale to China — critical minerals decoupling is now active statecraft, not theory. SCMP
• Research security reforms are tightening, but ASPI argues internal division in the