Imagine it’s 6 a.m. in Nairobi’s wholesale markets. Merchants deliver vegetables from farms to city retailers. Everyone uses M-Pesa, Safaricom’s mobile money platform. Suddenly, the system crashes, halting the market.
Across the continent in Accra, MTN’s MoMo also crashes, stopping business in Makola Market. With growing AI-driven cyberattacks, these scenarios are plausible. Safaricom, MTN, and the Kenyan and Ghanaian governments need an off switch to stop rogue AI agents from disrupting payment systems.
The Need for Local AI Control in Africa
The central issue is whether AI can fulfill its purpose while being controlled locally. The global AI debate often centers on China vs. the U.S., open vs. closed models, and dominance. In Africa, the focus shifts to local control and agency in building systems, even if they rely on foreign technology.
African countries are already building AI for digital transformation, as outlined in the African Union’s AI strategy. Policymakers are developing AI infrastructure but rarely discuss how AI drives safe economic growth. Developers across Africa are creating AI-powered innovations on the continent’s digital infrastructure, particularly mobile money.
Mobile Money: Africa’s Economic Backbone
In countries like Kenya and Ghana, mobile money is vital for daily economic activity. Small traders receive payments, households transfer funds, and bills are settled through these platforms. By 2025, over $2 trillion flowed through mobile money globally, with $1.4 trillion in sub-Saharan Africa. East Africa accounted for $806 billion, and West Africa for $498 billion.
A rogue AI agent disrupting transactions, manipulating accounts, or disabling access could severely impact livelihoods. Small-scale traders, who often live on daily earnings and lack credit or social welfare, would face devastating losses. The effects could extend to farmers and households dependent on the market.
Read Also: Non English languages cost more for AI
The answer lies in how Africa builds its AI systems, and who controls them. Whether a rogue agent can be stopped depends on the AI model’s deployment. If it runs on a centrally controlled cloud service, the provider may shut it down. If it’s an open-weight model deployed independently, containment becomes harder.
African developers are building largely on open models, which are arguably cheaper and can be run locally. However, these models offer less protection against misuse. Closed models have drawbacks too. When Hugging Face investigated a July intrusion, it used a Chinese open-weight model because U.S. models blocked requests with real attack code. Both open and closed models have vulnerabilities that Africa’s AI economy must address.
Africa must develop its own cybersecurity defenses for the AI age. This requires partnerships that position Africa as a technology developer, not just a consumer. These partnerships should ensure user control and avoid dependency on a single vendor. The goal is to create robust, secure applications using imported technologies to develop homegrown solutions.
African countries have limited cybersecurity capacity and rely on foreign providers. Key questions include whether African governments and firms have the monitoring systems, technical expertise, and redundancy to detect and contain rogue AI systems. Hugging Face, one of the world’s best-resourced AI companies, initially failed to escalate an intrusion alert. Few African institutions have similar capabilities.
Building these defenses requires a shift in African AI policy to prioritize resilience alongside adoption. Governments need human override mechanisms, segmentation, auditability, incident response, and offline fallbacks built into critical systems from the start. This is the off switch that must be in the hands of those responsible for protecting citizens and platforms when systems like M-Pesa and MoMo crash.
African countries currently have limited cybersecurity and incident-response capacity, relying heavily on foreign providers.