AI agents to drive fraud surge for Australian firms - SecurityBrief Australia
AI agents will drive a rise in fraud targeting Australian businesses in 2027, with organizations needing stronger data management and AI oversight amid rising scam and cybercrime losses.
Australians lost AUD $2.18 billion to scams in 2025, up 7.8%, while the average reported cybercrime cost for large businesses rose 219% to AUD $202,700. Matthew Lowe, Regional Director - Pacific at Anomali, warned that AI agents running entire fraud campaigns will be the major trend in 2027, as they draft lures, clone voices, cycle through stolen credentials and move money in minutes.
Fraud escalation
Fraud has become a central issue for security leaders and buyers, particularly in banking and government procurement. Lowe noted that "fraud tops the agenda for every Australian CISO I meet these days. It surfaces in every major enterprise and government tender we see, and tier one and tier two banks want to know exactly how deep our capability runs."
AI agents running entire fraud campaigns will be the major trend in 2027. We are already seeing this now, and it will escalate in 2027 as they draft the lures, clone the voices, cycle through stolen credentials and shuttle the money in minutes. A human still steers every campaign, but the agents never take holidays or sick leave. They rattle the door around the clock.
In 2027, expect deepfake executives to approve payments, AI-written emails to hijack supplier invoices, and synthetic identities to pose as real customers and staff. Payment redirection, investment scams and account takeover will all climb as criminals adapt to new security measures.
Target shifts
When the Scams Prevention Framework binds banks, telcos and digital platforms from 31 March 2027, criminals will migrate to softer targets: superannuation, real estate settlements, retail and small business supply chains. This shift will extend fraud risk beyond traditional payments teams into procurement, customer service, identity verification and operational technology.
Fragmented data tops the list of cybersecurity, privacy and governance roadblocks. Fraud signals scatter across email, identity, payments and endpoint tools, making it difficult to spot suspicious patterns before transactions are completed. Most businesses also struggle to distinguish between legitimate customers and AI agents acting on their behalf.
Defense strategies
Lowe recommended companies focus their fraud investment in three key areas: unifying security, identity and transaction data in one format; fusing threat intelligence into every event as it lands; and stretching identity controls to cover AI agents, monitoring their behavior and flagging synthetic ones before they transact.
AI will help combat fraud by helping defenders connect signals in seconds and intervene before money moves. It also frees analysts from chasing false alarms. The catch: AI inherits the flaws of the data it reads. Feed an agent fragmented data and it will make bad calls faster. Ground it in strong intelligence, govern every action it takes, and keep each decision explainable and reversible.
Fraud, security and identity teams should work more closely together because many fraud incidents begin with cyber events such as phishing attacks or voice cloning. Companies must also clean up data before introducing more automated systems and ensure every machine-made decision can be justified to regulators, boards and customers.
Lowe framed the challenge in terms of speed: "Fraudsters already move at machine speed. Australian businesses need governed decisions at the same pace." The growing automation of financial crime requires organizations to develop equally rapid response capabilities while maintaining strong governance and oversight.