Amnesty International hacked the mortgage verification system


Mortgage underwriting is run on documents. Payrolls, bank statements and tax returns are evidence that lenders check before approving a loan. Generative AI can now produce all three in formats that pass standard tests.

The Australian mortgage market could face up to A$4 billion (about $2.8 billion) in suspected fraud, as organized crime networks use artificial intelligence to fabricate financial documents underpinning home loan approvals, Broker Daily reported. I mentioned June 29, quoted from Australian Financial Review.

the Commonwealth Bank of Australia The Nightly alone is investigating up to A$1 billion (about $695 million) in potentially fraudulent loans I mentioned March 16. The fraudulent orders were mainly submitted by fake small business owners who used artificial intelligence to create fake accounts, earnings and invoices.

Financial crimes regulator Australia A. was held Vintel Alliance To allow banks to share intelligence with each other and with law enforcement, according to The Nightly.

document verification Not possible Save higher with Created by artificial intelligence Forgery

The Nightly report said some banks are now examining digital fingerprints of submitted documents, examining file data for evidence that the document was generated by artificial intelligence. This approach has limitations. The scammer, who produces AI-generated documents and then runs real payroll deposits through a real bank account for months before filing, has compiled a file that appears consistent on both sides of a verification check.

Some existing document verification tools may not be able to detect submissions that are AI-generated or digitally modified, according to Legal commentary From a law firm Minter Ellison.

He said the Australian case revealed a deeper failure in how the verification process was designed Dominic Taikoboss Thaddeus Martin ConsultingGallagher I mentioned May 19.

“We were checking documents when we should have been checking people,” Tyco said, according to the report.

National Australia Bank He called for the creation of a national strategy to combat economic crime on June 27 advertisementSaying that the efforts of individual banks are no longer sufficient.

“This is complex organized crime that spans across industries and borders,” NAB said in the announcement, adding: “This is a system-wide problem, and it requires a system-wide response.”

Simon ConstantCommissioner Australian Securities and Investments Commission (ASIC), issued Open letter May 8 Financial services license holders warn that artificial intelligence is reshaping the fraud threat.

“This is not a remote or hypothetical danger,” Constant wrote. “It’s here now, it’s evolving rapidly and it requires the attention of boards and executives.”

scam It works requires Transgression borrower Documents completely

the Mortgage and Finance Association of Australiathe Australian Banking Association Other sector groups petitioned the treasurer Jim Chalmers To expand Consumer data right To give lenders direct access to approval Australian Taxation Office Income statements and ASIC registration information, according to Broker Daily a report.

Verifying income directly against government records bypasses the entire document layer, the report said. The Federal Government has committed A$62 million (about US$43 million) over two years from 2026 to 2027 to fund the next phase of the consumer data right, including integrating tax authority data into the open banking ecosystem.

A borrower submitting a payslip provides a document that can be forged. A borrower whose income is confirmed directly from the tax authority’s payroll system provides data that cannot be verified.

“Digital finance is built around the assumption that more data creates more certainty,” PYMNTS I mentioned June 10. “Synthetic borrowers overturn this hypothesis.”

For all of our PYMNTS AI coverage, subscribe to our daily newsletter Amnesty International newsletter.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *