Menu close

Olga prevents fraud of DKK 200 million

Since April 2024, the vigilant and now two-year-old Olga has virtually stopped one of the most widespread fraud methods experienced by Vipps Mobilepay.

13. Jul 2026
3 min
English / Dansk

Previously, Norwegian users of Vipps, the Norwegian equivalent of Mobilepay, were defrauded of millions of kroner through social engineering.

In April 2024, fraud activities peaked at DKK 6.4 million as fraudsters all too often succeeded in scamming money from victims by, for example, pretending to be from the police or the bank. 

But in the same month, i.e. April 2024, Vipps plugged in a brand new AI model called Olga. And it has practically put a stop to this type of fraud since then.

During the first six months after the launch of Olga, the number of social engineering scams was reduced by 98 per cent. In March this year, it decreased even further to more than 99 per cent, and Vipps now barely has one fraud case of this type per month.

According to Vipps, a conservative estimate shows that Olga has deterred fraud activities of around DKK 200 million. The amount is based on calculations of what the situation would have looked like if fraudsters had continued as before, and Vipps' AI model had never been made.

"Vipps is a closed ecosystem as we know both the sender and the recipient. This gives us more context and more data points when Olga assesses payments. The ecosystem means that the AI model has good working conditions," explains Hege Hagen, Head of Financial Crime Prevention at Vipps, in an announcement.  (in danish)

(The article continues below)

"Olga strikes back" 

Vipps and Mobilepay merged in 2022 but operate with their own brands in Denmark and Norway. 

While vigilant Olga has celebrated her two-year anniversary in operation at Vipps, Mobilepay has used her since January. 

Olga analyses every single payment in real time and assesses the risk of fraudulent activity. The payment is cancelled if the risk is sufficiently high, and at the same time, Olga becomes better at recognising different fraud patterns.

Fraud is combatted specifically by analysing a number of measuring points, Jeanette Hertzum, Country Manager for Denmark at Vipps Mobilepay, told Finanswatch in January. (in danish).

Back then, she also revealed that Olga had reduced fraud by 50 per cent in the first month of use in Denmark.  

The good results have meant that the AI model has been continuously scaled and, today, Olga is protecting 'friend payments' in both Norway and Denmark. Recently, the system also began to stop fraud attempts among business payments. 

"Fraud patterns change rapidly, and we're therefore constantly working to improve Olga. As the model learns and encounters new threats, we develop new features that make it even more accurate," explains Hege Hagen.

The diligent AI model has not only been named Olga because it's a girl's name with Nordic origins. 

"The name stems from what is often coined 'Olga fraud', implying that fraudsters typically target the elderly and lure them into sending money. The name also serves as a kind of tribute to the victims and as a sign that 'Olga strikes back'. And she has definitely done that," says Hege Hagen.

 

The KPIs of Olga

  • 20 milliseconds: the time it takes Olga to determine whether fraud may be involved. 
  • Each decision is based on more than 30 signals from over 20 data sources.
  • 109: the number of new Olgas that have seen the light of day since the original.
  • 20+: the number of fraud attempts prevented by Olga on a daily basis.

Latest news