My deepfake fraud journey into insurance
I’ve spent my morning creating an insurance claim for damage to my car. But the damage isn’t real, nor is the cost estimate/invoice, and it isn’t even my car. I don’t think anyone will be able to tell.
The work I’m doing is actually to help an insurance firm improve their detection models as they fight against the rise in deepfake and shallowfake-enabled fraud. With new advanced and accessible AI tools it is easier than ever to fabricate convincing claims, and every insurance firm should be preparing themselves.
AI enabled fraud in Insurance
Insurance is undergoing a digital transformation and one area which has seen substantial change is claims processing. Automated systems make assessments quicker, and online platforms are more convenient for customers. However, these also bring an opportunity for potential fraudsters using the latest deepfake and shallow fake AI tools.
Deepfake Fraud
Leveraging advanced AI, deepfakes produce hyper-realistic but fake audio, video, images or other content. For instance, as noted in a blog by Zurich fraudsters have tried to use deepfake technology to simulate a car accident, including fake damage photos that seemed highly realistic.
Shallow fake Fraud
Unlike deepfakes, shallowfakes can involve simpler editing techniques—such as altering timestamps, editing documents, or adding fake scratches to vehicle photos. For example, a recent article in the Guardian highlighted a car insurance scam where fraudsters manipulated photos to add fake damage. These edits are easier to produce, yet they can be highly deceptive and effective in misleading claim assessments.
AI tools are becoming increasingly advanced and accessible, lowering the barriers for fraudsters to create convincing content that can be used in claims. Fraudsters don’t need to be experts in photo or video editing. They can easily access AI tools that will create photorealistic images, videos or documents based on simple text prompts.
Poacher turned gamekeeper
While AI can be a tool for potential fraudsters, AI also offers an important weapon in detecting and stopping them. Machine learning algorithms can analyse vast amounts of data to identify patterns indicative of fraud, and image and video forensics can detect anomalies in video or images files that are invisible to the human eye. Here are 3 things we’ve seen leading insurance firms do to reduce fraud levels:
- Partnerships with Insurtech Detection Solutions: Insurtech solutions are rapidly emerging which are capable of scrutinising digital media and detecting fraudulent claims. Forensic software can analyse metadata, detect signs of manipulation, and verify the authenticity of submitted evidence. Insurance firms must be able to rapidly assess Insurtech solutions to make sure they have the capability they need. NayaOne’s Sandbox Platform enables Insurance firms to run PoC’s in weeks so they can identify the right tech partner for them.
- Employee Training and Awareness: Firms are training their teams, giving them examples of highly sophisticated fraud examples, and access to AI tools to experiment with to know what to look for. NayaOne has large datasets of deepfake and shallow fake media to help firms train their staff, and to help them test potential partners' detection capabilities. An informed workforce is crucial in identifying and responding to suspicious activities promptly.
- Invest in Partnerships and Advanced Analytics: Partner with companies specialising in AI-driven analytics to develop models that can detect irregularities in claims data. Predictive analytics can flag high-risk claims for further investigation.
Conclusion
Digital tools are transforming insurance, offering both remarkable benefits and new challenges. While fraudsters are attempting to use AI to exploit these technologies, AI also holds the key to robust defence mechanisms. It is clear what insurers must do: leverage AI’s ability to analyse data for anomaly detection, foster partnerships with leading AI detection companies, and train staff to stay vigilant. In doing so, they can outpace fraudulent activities and minimise losses.
