On 1 July 2026, Australia’s anti-money-laundering regime stops being a banking-and-casino regime. From that date, AUSTRAC’s obligations extend to lawyers, accountants, conveyancers, real-estate professionals and dealers in precious metals and stones — the long-delayed “tranche 2” reforms that bring tens of thousands of first-time reporting entities under the AML/CTF Act. Enrolment with AUSTRAC opened on 31 March and closes 29 July.
The date is not arbitrary, and neither is the timing. It lands in the same quarter that financial-crime regulators across Australia and New Zealand are reporting the fastest investment-scam takedown pace on record — fraud that is now produced at industrial scale by generative AI. The same technology shows up on both sides of the ledger: it is the attack vector flooding Meta and search with fake ads, and it is the new compliance burden landing on professions that have never run a transaction-monitoring program.
The attack side scaled first
The clearest read on the attack volume comes from New Zealand. The Financial Markets Authority identified 110 scam ads in a single 24-hour sweep of Meta platforms, and has flagged more than 190 fake trading-platform websites for removal since the start of March 2026 (FMA). The mechanics are now formulaic: fake news articles carrying the logos of real outlets — RNZ, TVNZ, the NZ Herald — wrapped around deepfake videos of local politicians and business figures endorsing a platform. Victims register their details, get a call from a fake “broker,” and are walked into an initial deposit of around US$250.
Across the Tasman, the regulator’s language is blunter. ASIC Commissioner Alan Kirkland: “Scammers are using artificial intelligence to make fake investment ads look more polished, more convincing and harder to spot. We’re seeing AI being used to create professional videos, fake endorsements and targeted ads” (Finance Magnates). The volume is the point. Generating a thousand variant ads, each tuned to a different audience segment, used to be the bottleneck. It isn’t anymore.
The takedown numbers are not a measure of how much fraud exists. They are a measure of how fast the production line runs.
That distinction matters for what 1 July actually changes. Takedowns are downstream — they catch the ad after it ships. The reforms are an attempt to move enforcement upstream, to the point where dirty money touches a professional service.
Tranche 2 is a maturity problem, not a coverage problem
The policy logic of tranche 2 is sound and overdue. Lawyers, accountants and real-estate agents are the “gatekeeper” professions — the structuring, conveyancing and trust-account services that launder proceeds need a credentialed intermediary, and Australia has been an outlier among FATF members in leaving them outside the net. AUSTRAC’s own 2025-26 priorities name cash-intensive sectors and the misuse of AI to facilitate money laundering as focus areas, and external counsel reads “targeted, multi-sector enforcement” into that (A&O Shearman).
The problem is not the coverage. It is the maturity gap. A mid-tier bank has spent two decades building suspicious-matter reporting, customer due diligence and a compliance function with a named officer at management level. A suburban conveyancer or a two-partner accounting firm is standing all of that up from zero, with a July deadline. AUSTRAC has shipped “Program Starter Kits” precisely because the new cohort has no institutional muscle memory for any of it.
So the squeeze is asymmetric in a way the policy framing doesn’t capture. The threat these new entities are nominally meant to help catch is AI-industrialised, adaptive and operating at machine speed. The defenders are running their first-ever AML program on a starter kit. The mismatch is not a transition problem that resolves once everyone enrols. It is structural: the cheapest part of the laundering chain to automate is the front-end deception, and the most expensive part to stand up is human-supervised compliance at thousands of small firms.
The implication: detection is now an AI-versus-AI problem the newcomers can’t fight
The uncomfortable conclusion is that the obligation and the capacity to meet it are arriving on different timelines. Industrial fake-ad generation is, functionally, an AI detection problem — volume and variation defeat manual review, which is exactly why FMA is counting ads per 24 hours rather than per week. The reporting entities best placed to deploy AI-grade monitoring are the banks that already have it. The entities tranche 2 newly captures are the ones least able to.
That points to two near-term dynamics worth watching. First, a vendor wave: AML-tech and KYC providers are the obvious beneficiaries of a forced cohort of tens of thousands of buyers with a hard deadline and no in-house capability — Moody’s and others have been positioning for exactly this. Second, an enforcement-tone question. AUSTRAC has signalled intensified enforcement, but pursuing a first-year conveyancer for an imperfect program — while AI scam factories run at record pace one layer up — would be a poor use of a regulator’s scarce attention. Early enforcement will more plausibly target egregious non-enrolment and the genuinely complicit, not honest firms still climbing the curve.
The deeper point is that 1 July does not close the gap the FMA’s takedown numbers describe. It moves enforcement to the gatekeepers, which is the right place for laundering. But the fraud production line that motivated the urgency lives in a different layer — ad networks, deepfake tooling, platform moderation — that tranche 2 does not reach at all. New Zealand’s experience is the preview: the FMA can flag 190 sites since March and the next 190 are already generating. Australia is about to ask its smallest professional firms to help hold a line that AI keeps redrawing faster than anyone can staff against it.
The reforms are necessary. The expectation that they will, on their own, blunt an AI-scaled threat is the part that won’t survive contact with July.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.