JetherVerse News
AI· 2 min read

OVEX deploys AI fraud detection built by DataProphet

South African crypto platform OVEX launched a custom AI fraud system built by DataProphet. In testing the system caught 98 per cent of fraudsters, against 75 per cent for the old rules-based system.

By JetherVerse News

South African crypto platform OVEX has gone live with a custom AI fraud detection system. Disrupt Africa reported the launch on November 17, 2026. DataProphet, a machine learning consultancy, built the system, which caught 98 per cent of fraudsters in testing.

OVEX started in 2017 and lists South Africa's largest selection of digital asset markets. DataProphet formed in 2014 and builds custom machine learning for financial services, manufacturing and other sectors. The two companies began work in April to replace OVEX's rules-based fraud system with AI trained on how fraudsters behave.

The system does not judge an account from one snapshot. Instead, the system reads a user's full activity history in sequence, including every login, deposit, trade and withdrawal. Two models assess each account. One trains on fraudulent behaviour and the other trains on honest customers. An account leaning towards the fraud model goes to investigators.

By late July, the model caught 98 per cent of fraudsters in testing, against 75 per cent for the old system. The model also flagged 100 per cent of fraud cases never seen before. OVEX expects 99 per cent fewer false alarms and a 92 per cent reduction in undetected fraud. The platform includes a tool showing investigators when an account's behaviour turned suspicious, plus a dashboard for retraining the models.

Chief executive Jonathan Ovadia said DataProphet built a system tailored to the platform and gives OVEX a more sophisticated way to identify fraud as the methods change. DataProphet chief executive Ridwaan Seedat said applied AI does not have to be a black box or a heavy compliance burden.

The figures come from testing and company expectations, not from a full year of live results. Treat the 99 per cent and 92 per cent numbers as targets.

Nigerian fintechs and exchanges fight similar fraud. A sequence-based approach, using the full history of each account, gives a clear model to study.

Compare your own fraud rules against the history of each account, not a single transaction.

Source: Disrupt Africa

Get the news in your inbox

The tech stories that matter from JetherVerse News, with the source on every one.

Free. Unsubscribe any time.