On February 27, media reported that the Commonwealth Bank of Australia had self-reported concerns to police and corporate regulators about possible mortgage fraud estimated at approximately $1 billion.
Australia’s largest bank has not released details about the alleged fraud, but reports said the bank identified the problem with mortgage applications that were referred through both brokers and introducers.
The investigation into the bank, which reportedly began in July last year, intensified after rival Big Four financial institution National Australia Bank faced fraud charges of around $150 million related to an operation known as the Penthouse Syndicate.
The move comes amid growing concerns across the brokerage industry. Almost three-quarters of Australian mortgage brokers said they had been affected by fraud or fraud in the 12 months to September 2025, up from 26% in the same period last year, according to research by Equifax.
Although brokers are increasingly aware of the risks, identifying fraudulent documents can still be difficult.
Fini Mortgages principal Eva Roisance told Broker Daily Uncut that obvious discrepancies can often be identified through standard testing, but increasingly sophisticated counterfeits can be much harder to detect.
“I’ve seen very obvious mistakes where the numbers simply don’t add up and all it takes is a few checks to know something is wrong,” Roissance said.
“But now if we’re good enough to make sure everything meets the criteria, how do we know it’s fake?”
Technology that drives both fraud and detection
In the case of CBA, there were reports that AI may have been used to forge documents. As technology advances, fraud techniques and fraud detection methods are also rapidly evolving.
Brett Spencer, president of the Australian Financial Brokers Association and founder and CEO of AI-driven document and data verification platform DocuScan, said fraudsters were increasingly using widely available AI tools to create convincing documents.
“Frauds are 100 percent more sophisticated,” Spencer said.
“It’s often very community-driven, using general-purpose AI technology to create documents that look and feel real, but aren’t real.”
One of the methods used to detect fraudulent documents involves analyzing metadata, the underlying data that describes how a document was created or modified.
“There are various websites that create fake bank statements and other documents that are pretty convincing, but when you look at the metadata layer of that document, the metadata doesn’t match, which makes this type of forgery a very blatant tool,” Spencer said.
AI can also analyze data points in documents at scale and identify discrepancies that may not be immediately apparent to human reviewers.
“What an AI platform can do is cross-reference bank account details on payslips, validate calculations, and verify dates and data,” Spencer says.
“You can also go beyond the typical 90-day period to determine whether the income is short-term or ongoing.”
DocuScan, which uses AI to process and extract information from financial documents, also develops tools designed for brokers.
The company announced the launch of two broker-focused platforms: ComplyX and FraudX. This allows brokers to upload client files for automated compliance, fraud, and validation checks.
At the aggregator level, AI is increasingly being used to identify patterns of suspicious behavior across loan applications.
Shirley Elliott, head of compliance at AFG, said the technology has significantly improved an organization’s ability to detect fraud trends.
“AI has proven to be truly useful in identifying trends in fraud across large volumes of applications, but until now it simply hasn’t been scalable,” Elliott said.
“Where this brings the most value is through pattern recognition, flagging applications that share characteristics with previously substantiated fraud cases and escalating those cases for further investigation.”
She added that new machine learning tools can analyze entire groups of applications at once to identify suspicious patterns.
“What’s even more interesting is where this is going. We’re now seeing machine learning models that not only evaluate applications individually, but also look across queues and surface suspicious patterns, such as clusters of applications sharing IP addresses or suspiciously similar email formats,” Elliott said.
“This is a meaningful change in capabilities.
“Another particularly valuable advancement is the advent of fraud detection software that streamlines the process of verifying documents submitted by customers for loan applications. Automating these checks allows lenders and mortgage originators to efficiently identify potential fraud, reducing manual effort and improving valuation accuracy.”
Human oversight remains essential
Despite the increasing use of AI in fraud detection, Elliott cautioned brokers not to rely solely on automated tools.
“Brokers need to be careful not to delegate decisions to AI,” she says.
“Under the NCCP Act, there is a clear duty to ensure that the information you rely on is accurate and free of anomalies.
“The AI that brokers currently use, at least, can’t see everything. Payroll staging is a great example of the kind of fraud that doesn’t show up on automated checks. It still requires a human eye and the right questions.”
Spencer similarly emphasized the need for human oversight alongside automated verification tools.
“Decisions still require a human-involved process,” he said.
“AI can help with fraud detection, document validation, data validation, but someone still needs to look at something and say, ‘That doesn’t seem right.’”
[Related: Threat actors pull back on publishing stolen youX borrower data]
