Bank bosses are generally among the most enthusiastic artificial intelligence evangelists in the corporate world. AI can not only cut costs, they claim, it can also improve customer service and even protect people from scammers and others who are up to no good.
Time will tell if AI lives up to all this hype. What has undoubtedly become clearer, however, is that in financial services, AI also has a more sinister side: it is extremely useful for people trying to cheat the system.
The latest example of this is AI-facilitated mortgage fraud. We learned last week there could be hundreds of millions of dollars in fraudulent loans held by Australian banks, and AI appears to have accelerated this problem.
The financial crimes watchdog, Austrac, revealed that after conducting a probe together with banks dubbed “Operation Claw”, it found there were “potentially hundreds of millions of dollars in suspected fraudulent loans”. The investigation was carried out by the Fintel Alliance, which includes the biggest banks in the country, and it drew on data from 10 major banks.
The suspected fraud was enabled by fake payslips, fake company documents, and more. This array of fakes – which banks say can be much more easily created using AI – was then used to support home loans which were mostly linked to houses in Sydney.
In one sense, this is not a new problem for banks, which have always had to contend with fraud. But in another, Austrac’s finding underlines how AI has amplified problems such as scams, hacks and fraud.
It shows the need for ongoing technological change in the industry that is already one of the biggest investors in technology in the country. And it also suggests we need to make bigger policy changes in how our data is shared and verified in the financial system – more on that later.
So, how big a problem is mortgage fraud for banks? And what can they do about it?
Austrac is the most feared regulator in corporate Australia because of its ability to seek gargantuan fines for breaches of anti-money-laundering and counter-terrorism financing laws. So when Austrac talks, the industry listens. Yet, it’s important to note that last week’s update on potential mortgage fraud did not find evidence of widespread money laundering. Hefty fines are not in play at the moment.
What it did find were weaknesses in banks’ systems that could be exploited by criminals. The basic problem is that banks failed to detect false documents that were used to support loan applications – and it’s become much easier to generate such fakes.
“The scale of this activity should be a wake-up call for every lender. The same warning signs were found across banks that together cover the vast majority of Australia’s mortgage market,” Austrac’s boss Brendan Thomas said.
Importantly, it’s not only banks in the spotlight: the names of mortgage brokers, lawyers and accountants who may have been involved have been given to regulators and the police.
Yet, it is clear that banks need to get better at spotting fake payslips, among other changes, and AI technology will surely play a role here.
Jarden analyst Matthew Wilson says Austrac’s findings suggest criminals have found ways to exploit AI faster than banks have managed to bolster their defences. Wilson says that Austrac’s detection of “hundreds of millions of dollars” in suspect loans is a small share of the mortgage market, which is worth about $2.6 trillion.
Banks have long argued for a change in our antiquated reliance on payslips and other documents to prove a borrowers’ income.
However, the worry is that it suggests sophisticated criminal syndicates have found a way to “wash” dirty money through the banking system – similarly to how criminals have washed money in casinos.
“There’s no risk of financial loss for the banks, it’s just that their platforms are being used for illicit activity,” Wilson says.
“What’s happened is the criminals have evolved faster than the banks. The criminals are using AI better than the banks. The banks have got to be better at using technology to detect the fraudulent applications. From a money-laundering perspective, it’s critical.”
When Commonwealth Bank chief Matt Comyn was asked about the problem at the bank’s recent results release, he stressed the bank had not found any evidence of organised crime involvement or professional money laundering. But he said it continued to be “an area of focus for us”. Comyn also pointed to significant changes in the “risk landscape” in areas such as cybersecurity, scams, fraud and financial crime.
To be fair, it is not only up to banks to fix this. The government can also help.
Banks have long argued for a change in our antiquated reliance on payslips and other documents to prove a borrowers’ income, because of the risk that such documents are faked.
Instead of relying on documents, the banks want the government to allow the Tax Office to securely share customers’ data, as a way of proving their income. Customers would have to consent to their tax data being shared – something that could occur through a system known as the consumer data right.
The government committed to exploring the idea in the budget, and Australian Banking Association chief executive Simon Birmingham last month called for change “as soon as possible”. He called the reliance on payslips and tax returns “an anachronism in the digital age” that’s created unnecessary risk for the financial system.
“Fraudulent loan documentation is a growing problem, and artificial intelligence is making fake payslips and doctored statements frighteningly easy to produce. Verified ATO data would give lenders a single, trusted source of truth for a customer’s income – helping to prevent document fraud at its source,” Birmingham said.
Making this happen won’t be simple. Borrower privacy must be protected, and the consumer data right has been slow to take off in other parts of banking. The budget committed $62 million over two years for improving the consumer data right, including looking at system for sharing Australian Taxation Office data.
So, it is a fair bet that people applying for loans will need to keep producing payslips for a while yet. And until there is a more sophisticated way of checking customers’ income, banks will need to get better at spotting the fakes that are being churned out by AI.
The Business Briefing newsletter delivers major stories, exclusive coverage and expert opinion. Sign up to get it every weekday morning.
Clancy Yeates is deputy business editor. He has covered banking and financial services, and was previously national business correspondent in the Canberra bureau.Connect via X or email.