The $4 Billion Syndicate: AI-Forged Loans and Australia's Mortgage Fraud Crisis

A former banker walked into a luxury Sydney property settlement knowing the loan he arranged was built on fabricated income statements. His clients - allegedly part of a criminal network washing drug trafficking proceeds through real estate - were never going to repay it. They did not need to. The property was the point.

That mechanism sits at the heart of what is now Australia’s largest active fraud investigation. The probe covers approximately $4 billion in home purchases, with suspected fraudulent loans held by National Australia Bank, the Commonwealth Bank of Australia, Westpac, and ANZ. NSW Police Financial Crimes Squad is investigating. ASIC has declined to comment. The banking sector is reviewing its loan books.

What the Penthouse Syndicate actually did

The alleged ringleader is Andrew Hu, a former CBA and NAB banker whose insider knowledge of lender processes reportedly became a blueprint for organised crime. The network he allegedly built - known to investigators as the Penthouse Syndicate - recruited lenders, mortgage brokers, real estate agents, and lawyers who knew exactly which checks to satisfy and which documents to forge.

NAB’s initial estimate of its exposure to the Syndicate was around $150 million. That figure has grown as the investigation expanded to include the other three major banks. CBA, which first notified police in late 2025 after a whistleblower complaint in February that year, has identified approximately $1 billion in suspected illegitimate loan approvals and is reviewing further.

The criminal goal was not just obtaining credit. The loans were the mechanism for laundering proceeds from drug trafficking and other organised crime activity - moving illicit cash into residential property, one of the hardest asset classes to trace. An offshore criminal network with liquid proceeds and a compliant broker can buy a $2 million property, hold it for two years, sell it at market rate, and emerge with clean capital and a plausible transaction history. The property market becomes the washing machine.

How AI collapsed the fraud barrier

Mortgage document fraud is not new. What is new is the rate at which it can now be produced at scale.

Forging a payslip used to require design skill, familiarity with a specific employer’s template, and access to software most forgers did not have. Tax returns required ATO-format expertise. Income statements needed the right superannuation contribution rates, valid ABNs, and accurate PAYG codes. The fraud was limited by the effort it demanded.

Generative AI removed that constraint. Modern language and document generation models can produce a convincing payslip, bank statement, or income statement in seconds. The output can be tailored to a named employer’s template, carry a valid ABN, show the right pay cycle, and format salary figures to look like legitimate PAYG income. Security researcher Jamieson O’Reilly described the Penthouse case as potentially “the first at this scale where AI-generated documentation appears to have been a primary mechanism.”

That assessment fits what CrowdStrike’s 2026 Threat Hunting Report found across the financial sector more broadly. Intrusion activity against financial institutions rose 11% year on year. Nation-state intrusions targeting the sector surged 29%. AI-enabled adversary activity across all sectors surged 89% in 2025. The report’s framing: “Frontier AI is collapsing the window between vulnerability discovery and exploitation.” In mortgage fraud, that window collapsed into a matter of prompts.

Who is exposed beyond the banks

The banks are the headline. They are not the only parties left holding risk.

Mortgage brokers who submitted fraudulent applications - whether knowingly or because they accepted documents at face value - face professional indemnity exposure and potential criminal referral. The Penthouse Syndicate is alleged to have recruited insiders, but the investigation’s scope also covers brokers who may have been deceived rather than complicit. The distinction matters for liability. It does not make the outcome different for the lender.

Real estate agents and conveyancers are also in scope. A conveyancer processing a settlement receives payment instructions from a buyer whose identity documents, loan approval, and source-of-funds documentation may all be AI-generated. Without a live verification check against the ASIC registry and a bank account verification step, the conveyancer has no mechanism to confirm the settling party is who they claim to be.

The property purchases in this investigation moved through normal settlement channels. The fraudulent nature of the underlying loans was invisible at the point of settlement. That is the structural problem - and it is one that is not solved by checking the documents more carefully, because the documents are the problem.

Where the verification gap actually sits

Standard mortgage approval processes check documents. They do not verify the entity behind them.

A submitted payslip says a person earns $160,000 a year at a named employer. A lender’s processing system confirms the document is formatted correctly, the ABN resolves, and the figures add up. What it does not do - unless the lender has built a specific live integration - is confirm the person actually works there, that the bank account their salary is paid into is genuinely theirs, or that the employer is a real operating business rather than a shell company registered three months ago.

The Penthouse Syndicate exploited that gap precisely. Insider brokers and lenders knew the specific triggers for manual review and kept submissions calibrated to avoid them. AI-generated documents that pass optical inspection do not trigger document-level fraud alerts. The gap between visual review and live data verification is where the $4 billion sat undetected.

Industry bodies have now asked government to expand the Consumer Data Right (CDR) to include ATO income data and ASIC registry information. The logic is sound: if a lender calls an authoritative data source directly rather than accepting a submitted document, the AI document forgery attack fails at its source. There is no payslip that overrides a live ATO income confirmation. There is no forged company record that survives a real-time ASIC registry match.

That CDR expansion has not happened yet.

What regulators are asking for right now

AUSTRAC chief Brendan Thomas acknowledged that AI “creates both risks and opportunities” for financial crime detection. APRA has gone further, calling for a major uplift in AI governance and risk management across the banking sector - specifically in loan origination and fraud detection.

NAB’s public position is that industry effort alone is not sufficient. The bank has called for a National Economic Crime Strategy to enable coordinated intelligence sharing, cross-sector collaboration, and consistent regulatory approaches. NAB has suspended or exited multiple parties, referred cases to authorities, and is working with the Fintel Alliance and law enforcement agencies. The bank’s statement to the industry was direct: “This is complex, organised crime that spans industries and borders.”

ASIC has declined to comment on enforcement intentions. NSW Police Financial Crimes Squad is the active investigative authority.

What is notable is the absence of a single coordinating body. AUSTRAC is monitoring for AML risk. APRA is pushing AI governance. The industry is requesting CDR access. Police are pursuing individual cases. The Penthouse investigation is a multi-year affair. In that gap, the next syndicate is already setting up.

What mortgage and settlement professionals can do now

The CDR expansion will take time to legislate and deploy. The investigation will take years to resolve. Neither timeline helps a conveyancer settling a transaction next week.

Verification steps that exist today do not require new legislation.

Verifying the bank account that will receive settlement funds is a live check available now. An account name match against the ASIC company register for business settlements, or against identity documents for individuals, flags a shell entity or a name mismatch at the point of settlement - before funds move. That check does not require ATO data or CDR access.

Verifying the person behind the transaction matters equally. A live check against the ASIC registry, combined with identity document verification that confirms liveness and document authenticity rather than just visual inspection, closes the identity fabrication vector the Penthouse Syndicate appears to have relied on. A liveness check defeats a scanned AI-generated document. A registry match defeats a shell company presenting as a legitimate buyer.

The Syndicate’s alleged durability rested on insiders knowing which checks were automated and which were eyeballed by a human reviewer. A verification layer that operates independently of the loan origination team - checking the person, the business, and the bank account against live authoritative sources - is not something insiders can calibrate around, because the check is not within their process to influence.

The documents in this case looked correct. The fraud was in what the documents were not verified against. That is the lesson - and it does not require waiting for the regulator to act.

---

ezyshield is on a mission to eliminate payment fraud in Australia. We verify the person, the business, and the bank account before money moves - so payments only ever land with who they’re meant to. Learn more