F.A.Q.
What is DAFNE, in simple words?
DAFNE is VISADA’s artificial intelligence system that analyzes accident images to identify recurring damage and potential fraud.
It does on a large scale what today only an expert expert can recognize on a case-by-case basis: it compares the damage, correlates it with the historical data and produces objective and measurable evidence.
Does DAFNE replace experts or fraudsters?
No. DAFNE does not replace experts or fraud prevention.
It is a solution designed for insurance companies that supports internal teams, providing evidence and reports that experts can use to make faster and more informed decisions.
What kind of fraud can it detect?
DAFNE analyzes accident images to identify repeated, reused, or inconsistent damage between different claims, even over time or on different vehicles.
It is particularly effective in cases where the damage is recognizable to a human expert, but impossible to manually intercept on large volumes.
In particular, DAFNE identifies:
damages not attributable to the declared accident, with reused or slightly modified images
reassembly of already damaged parts on different vehicles
image manipulations (cropping, copying, recurring elements)
the same vehicle presented with different license plates, even foreign ones
In these cases, DAFNE transforms seemingly legitimate images into comparable and verifiable evidence.
If damage is visible to the human eye, why do we need AI?
Because the difficulty is not seeing the damage, but remembering it, comparing it, and connecting it on a large scale. An anti-fraud team can identify an anomaly on a single claim. DAFNE does this across tens of thousands of claims, consistently and continuously, tirelessly, without bias, and without losing historical memory.
Does DAFNE give an automatic verdict of fraud?
No. DAFNE does not issue verdicts or make automatic decisions. It produces reasoned and traceable reports based on objective and verifiable evidence, understandable even by non-technical specialists. This makes the decision-making process faster and more effective and facilitates the fraud prevention process within the insurance company’s anti-fraud processes.
How reliable is the system?
DAFNE is trained on large volumes of real-world data and is constantly monitored. Each report is verified by an expert operator before being delivered: this ensures that, even in the presence of AI errors, the output remains reliable, verifiable, and defensible, even during internal audits or reviews.
How long does it take to start seeing results?
Results come immediately. At the start of the collaboration, the insurance company provides a history of claims images (typically from the last 2–5 years). This history is analyzed and indexed by DAFNE within a few days.
This way, each new claim is immediately compared with the historical data, making the system effective from the very first stages of use. Over time, its value grows further as the number of cases analyzed and relationships identified increases.
Is the use of DAFNE compliant with GDPR and privacy regulations?
Yes. DAFNE is designed to operate in compliance with current regulations, with particular attention to data minimization, security, traceability, and process auditability.
By design, we remove identifying elements such as license plates and faces from the images, because the analysis focuses exclusively on the damage and not the subject.
From a procedural standpoint, at the start of the collaboration we propose a DPIA (Data Protection Impact Assessment) that already includes the DAFNE use case, to support the company in correctly framing the system in accordance with the regulatory framework.
How is company value measured?
Value is measured in reduced fraud costs, increased efficiency of anti-fraud teams, and better resource allocation.
DAFNE also helps speed up pre-settlement processes, reducing the overall risk on the claims portfolio.
Furthermore, the system’s output provides more robust evidentiary evidence, which can be used both to support the rejection of a compensation claim and, when necessary, in dispute or legal proceedings.
Each project is accompanied by clear and verifiable metrics, shared with the company.
Why should this also concern insured citizens?
Because insurance fraud has a real cost that falls on everyone, in the form of higher premiums.
Making anti-fraud analysis more effective means reducing waste, limiting incorrect behavior, and contributing to a fairer insurance system, where costs are not passed on to those who behave correctly.
