Starling Bank has unveiled a new AI-powered scam detection tool, making it the first UK bank to embed this kind of technology directly into a retail banking app. The feature, called Scam Intelligence, is designed to help customers identify and avoid online purchase scams before they send money.

Targeting the growing threat of fraud on online marketplaces, the tool lets users upload screenshots or images of items, listings, and chat conversations with sellers. Starling’s system then uses a combination of image analysis, natural language processing, and behavioural indicators to flag potential risks, such as pressure to pay outside secure platforms, generic or low-quality product images, or suspicious seller behaviour. You can see more detail in Starling’s official release and coverage from Yahoo Finance UK and The Fintech Times.

The model is powered by Google Cloud’s AI infrastructure, using the Gemini family of large language models, and runs within Starling’s secure cloud environment. The bank states that user data is not used to train external models, which aligns with the privacy notes in its press announcement.

Tackling a billion-pound problem

Fraud is one of the UK’s most damaging and fastest-growing financial crimes. According to UK Finance’s Annual Fraud Report 2025, total fraud losses remained at about £1.17 billion in 2024, and authorised push payment (APP) fraud losses were just over £450 million. Purchase scams on online marketplaces remain a persistent vector. Starling’s tool aims to intervene before a payment is made, shifting from reactive reimbursement to proactive prevention.

How it works

• Users upload an image or message thread from a suspected scam, for example a marketplace listing or DM.
• The AI scans for common red flags, including unusually low prices, requests to bypass platform payments, pressure to pay quickly, and mismatches between item descriptions and photos, as outlined in Starling’s launch notes.
• The tool then provides a real-time risk assessment so customers can reconsider or cancel the transaction before sending funds.

Adoption and implications

The feature is currently opt-in, consistent with Starling’s approach to privacy safeguards described in the press release. It complements existing real-time payment monitoring and fraud alerts, and reflects a wider trend of banks embedding consumer-facing AI features to improve protection. Government voices have welcomed this direction; press coverage notes praise from the UK’s Minister for Fraud regarding the potential of AI to improve customer protection and outsmart scammers, alongside sector calls for broader collaboration across platforms and telecoms, as seen in recent reporting by the Financial Times.

A few open questions

Despite the promise, long-term success depends on real-world effectiveness. Will customers use the tool proactively, can the system keep pace with evolving tactics, and how should banks balance privacy, usability, and risk as AI scales? For consumers, official guidance on reporting and support remains essential, including the ability to report incidents to Action Fraud and to review the broader fraud trends highlighted by UK Finance.

For now, Starling has taken a notable step toward putting more power into customers’ hands, offering a practical second layer of due diligence inside the app.

‍