The mortgage industry is seeing emerging fraud risks posed by AI after a major lender detected a suspicious case involving an AI agent.
Brightstar Group's chief executive, Rob Jupp, revealed the incident during his closing remarks at the lender's Specialist Lending Expo last week, as he warned the industry against focusing solely on the potential benefits of AI. He said one of the biggest concerns among major lenders was the threat AI could pose both to existing business models and through its potential use in fraud.
Jupp said: “AI seems all ‘upside at the moment’ but advisers shouldn’t forget who’s advice and ultimately, future is on the line, and its isn’t an AI agent, but the regulated mortgage adviser.
“The first piece of formal AI fraud in the UK in the mortgage market has probably already been done, and certainly the first multiple detections of an AI agent getting really close has just been frequently spoken about.”
He urged brokers and lenders to recognise the risks accompanying the rapid adoption of the technology, but also noted that AI was also creating significant opportunities for mortgage businesses.
He revealed that six months ago virtually none of Brightstar Group’s leads came through AI, compared with just under 5% today.
Jupp described the technology as potentially a “fantastic lead generator”, but stressed that its effectiveness depended on the quality of information feeding it.
His comments followed an earlier panel discussion at the Expo in which speakers highlighted the importance of maintaining human oversight of AI-generated information and decisions.
Mainstream lenders will use AI tools to pursue customers directly
The speakers agreed that AI will enhance rather than replace human decision-making in specialist lending, where complex cases will continue to require human judgement.
Hope Capital's chief financial and operating officer, Kate Cowan, said technology was helping the lender improve processes but could not replace human expertise.
She said: “We use technology to enhance what we can do, rather than actually do the job for us. It’s not where it needs to be where you can say, ‘I can replace that complete function or that complete person’.
“Every deal is different. Every deal is unique,” she said. “That’s where technology doesn’t replace us.”
She warned that over-reliance on generative AI could also create additional work and risk, particularly if users stop properly interrogating its output. Cowan said tools such as ChatGPT, Claude and Copilot could generate impressive-looking communications but sometimes overcomplicated relatively straightforward issues.
She added: “People will become lazier. Loads of people have the plugins for Claude or Copilot on the side of their emails, and they’re generating all these emails, but actually, are they reading them?
“Is what they say what they actually think, or is it what the computer thinks it wants to think? That’s going to cause issues further down the line.”
James Tucker, chief executive of Twenty7Tec, agreed human involvement remained crucial, pointing out that even as his firm's use of AI had boosted development productivity, outputs still needed validating.
However, he predicted AI could pose a greater challenge to advisers operating in the mainstream mortgage market.
He said there was “no question” that mainstream lenders would use AI tools to pursue customers directly, stating: “Mainstream lenders are going to go after customers in that way via these tools, and they will be successful as a result of it.”
Tucker predicts adviser share of straightforward “black and white" mortgage cases would decline as lenders push further into direct distribution.
But he said the same threat was less pronounced in specialist lending because of its complexity, grey areas and the difficulty of translating lending appetite into simple rules.
