Starling Bank Deploys AI to Combat Romance Scams and Break Fraudsters' Psychological Spells

Quick Summary

Starling Bank has launched an AI-powered romance scam detection feature within its app to identify fraudulent behavior and prevent financial loss. By using agentic AI to ask targeted questions, the tool breaks the psychological manipulation used by scammers, protecting five million customers from romance, investment, and deepfake fraud.

How does the new AI-powered romance scam detection protect users?

AI-powered romance scam detection works by intervening during the payment process when the system identifies high-risk transaction patterns. Instead of a simple warning, the Starling Assistant engages the user in a natural language conversation. It asks probing, objective questions about the recipient and the nature of the relationship to highlight inconsistencies in the scammer's story.

• Real-time intervention during high-stakes transfers.
• Behavioral change coaching to alert users to red flags.
• Direct support links to Starling's specialized fraud team.

The tool is specifically designed to counteract psychological manipulation, providing an objective voice that friends or family members may struggle to provide without causing conflict.

What types of fraud can the Starling Assistant identify?

Beyond romance fraud, the AI-powered romance scam detection engine is trained to recognize over 10 distinct scam categories. This broad coverage is essential as UK Finance reports that APP fraud losses reached £576.4m last year. The system effectively identifies investment heists, pension scams, and deepfake phishing attempts that are becoming increasingly sophisticated.

• 40% increase in investment fraud losses addressed.
• 300% improvement in payment cancellation rates for marketplace scams.
• Secure data processing using Google Cloud’s Gemini model.

By front-loading risk assessment, Starling ensures that customers are prompted to think critically before money leaves their account, addressing the 19% rise in total fraud losses.

How was the Tinder Swindler's experience used in development?

Starling Bank collaborated with fraud advocate Cecilie Fjellhøy to ensure the AI understood the emotional complexity of fraud. Fjellhøy, who famously lost nearly £200,000 to a romance scammer, provided insights into how scammers isolate victims from their support networks. This collaboration led to an AI that acts as a neutral third party, capable of asking the difficult questions necessary to break a scammer's hold.

• Expert-led design based on real-world victim experiences.
• Ministerial endorsement from UK National Fraud Minister, Lord Hanson.
• Agentic-AI framework that simulates human-like investigative reasoning.

FF NEWS TAKE:

Starling Bank continues to lead the UK market in proactive consumer protection. While many banks rely on static warnings, Starling’s use of AI-powered romance scam detection to create a conversational intervention is a genuine needle-mover. By addressing the psychological aspect of fraud—the "spell"—rather than just the transaction data, they are tackling the root cause of APP fraud. This is a benchmark for how agentic AI should be deployed in retail banking.