MTN is using AI to hunt down South Africa’s gigabyte thieves

MTN South Africa is using artificial intelligence to identify revenue leakage on its mobile network, including a form of abuse that turns zero-rated websites into gateways for data that should normally be billed.

The operator’s South African chief executive, Ferdi Moolman, told ITWeb that AI is increasingly being applied to practical network and commercial problems. One of those is revenue assurance: finding patterns that suggest customers or intermediaries are consuming data without the network charging for it correctly.

The example MTN highlighted is tunnelling through zero-rated websites. Universities and other organisations can arrange for selected sites to be accessible without consuming a user’s mobile-data allocation. That is useful for students and public-interest services, but it also creates an attack surface.

Fraudsters can attempt to disguise unrelated traffic so that it appears to be travelling to an approved zero-rated destination. If the technique works, the user can reach services that would normally consume paid data while the operator sees traffic associated with a free site.

AI moves into the network control room

This is a different AI story from the consumer chatbots that dominate technology headlines. Telecom networks generate enormous volumes of operational and billing data, making them well suited to machine-learning systems that can identify unusual patterns faster than manual investigations.

For MTN, the commercial incentive is straightforward. Mobile data is a core revenue stream, and even small leakages can become material when they are repeated across a network serving millions of customers.

AI can also help operators distinguish between a genuine spike in traffic and behaviour that looks engineered to exploit billing rules. That does not remove the need for human investigation, but it can narrow the field and surface suspicious activity sooner.

MTN says it is also applying AI to traffic analysis and other operational areas. The broader direction is towards networks that continuously analyse their own performance, demand and anomalies rather than relying only on fixed thresholds and periodic reports.

Zero-rating still matters

The challenge is that zero-rating has legitimate value in South Africa. Universities, education platforms and selected public services can use it to give people access when they have little or no prepaid data available.

A blunt response to abuse could therefore undermine the reason the programmes exist. The more useful approach is to detect the tunnelling behaviour itself while keeping legitimate zero-rated access intact.

That makes MTN’s deployment a useful test of AI in a less visible part of the telecom stack. The technology is not writing marketing copy or answering customer questions. It is being used to protect the economics of the network while preserving services designed to reduce the cost of access.

For South African mobile users, the immediate effect may be difficult to see. For operators, however, AI-driven revenue assurance could become one of the more financially important applications of machine learning on the network.