Australia’s mortgage market may be facing up to $4 billion AUD (about $2.8 billion) in suspected fraud, and the weapon is generative AI that produces fake payslips, bank statements and tax returns clean enough to clear standard checks. Commonwealth Bank alone is investigating up to $1 billion AUD ($695 million) in loans tied to fake small-business owners who used AI to fabricate accounts and invoices.
Here’s the part that should scare every fintech founder building document-verification tooling: your product might already be obsolete. Fraudsters aren’t just faking documents, they’re running real salary deposits through real accounts for months first, so both sides of the file check out. That’s not a detection gap you patch with a better OCR model.
ASIC commissioner Simone Constant told licensees in a May 8 open letter this “is not a distant or hypothetical risk.” NAB went further, calling for a National Economic Crime Strategy because single-bank fraud teams can’t out-engineer organized crime networks alone.
The real fix ditches documents entirely. Industry groups petitioned Treasurer Jim Chalmers to expand the Consumer Data Right so lenders pull income straight from the Australian Taxation Office, and Canberra committed 62 million AUD ($43 million) over two years to build it. That’s the tactical takeaway if you sell into lending: verified-source data integrations are about to matter more than your fraud-detection model, because a document is fake-able and a live government API call isn’t. Budget engineering time for CDR-style integrations now, not after your bank’s first eight-figure write-off.
Nathan Zakhary