Starling Bank has launched a new artificial intelligence tool designed to help customers identify and avoid romance scams, investment fraud, and deepfake phishing attempts. The Scam Intelligence AI-agent, now integrated into the bank’s Starling Assistant, became available to its five million customers on 24 June 2026.
The rollout comes as fraud losses continue to climb. UK Finance figures show that authorised push payment (APP) fraud cost Britons £576.4 million in the previous year, a 19% increase.
Within that total, romance fraud losses rose 23%, while investment fraud losses jumped a striking 40%.
The new feature aims to disrupt the emotional tactics scammers use to build false trust and pressure victims into sending money.
Starling Assistant, the bank’s conversational AI tool that handles everyday banking questions through voice or natural language, now includes specialised scam-detection capabilities.
When a customer describes a suspicious transaction, the AI can flag warning signs and guide them through a series of clarifying questions.
For example, if someone mentions transferring £3,000 for a new partner’s plane ticket, the assistant may ask how the relationship began, how long it has lasted, and why the partner cannot cover the cost themselves.
It then assesses the likelihood of fraud and recommends speaking with the bank’s support team before proceeding.
This latest development expands on Scam Intelligence technology Starling first introduced in 2025 to spot fraudulent online marketplace listings.
That earlier version already increased the rate at which customers cancelled risky payments by 300%.
The romance-scam detection elements were shaped with input from Cecilie Fjellhøy, a leading fraud advocate and the subject of the Netflix documentary The Tinder Swindler.
She lost nearly £200,000 after a scammer convinced her to take out loans and credit cards in her name.
Her insights helped highlight how fraudsters isolate victims from friends and family, making an objective third party like the bank especially valuable in revealing deceptive behavior.
Bernadette Smith, Starling’s Chief Customer and Banking Officer, stressed that banks must give customers practical tools to defend themselves when social media platforms fail to stop scams at source.
Harriet Rees, the bank’s Chief Information Officer, said the AI encourages people to pause and question their spending plans, helping them recognise that anyone can become a target.
The initiative has been welcomed by Lord Hanson, the UK Minister for Fraud, who noted the government’s £250 million investment in its new fraud strategy to tackle these crimes.
The Scam Intelligence tool runs on Google Cloud’s Gemini model. Customers must opt in to use it, and all data stays within Starling’s secure environment without being used to train external models. By combining real-time risk analysis with supportive questioning, Starling aims to give customers the confidence to resist sophisticated manipulation and protect their money from increasingly common digital threats.
