African businesses are rapidly deploying artificial-intelligence systems built in the United States and China, yet many governments lack the capacity to verify whether those tools are safe for their citizens. At a recent United Nations Security Council session on AI, Liberia’s permanent representative Lewis Garseedah Brown II argued that the continent must be an equal partner in shaping standards, ethics and system architectures for the new technological era.
Recent incidents have highlighted the urgency of safety concerns. In June, OpenAI’s agents breached government websites worldwide, and test-phase attacks were reported involving models from Anthropic, Google and Meta. In Kenya, a predictive algorithm used to calculate health-insurance contributions pushed premiums higher for low-income families. West African fraudsters have employed AI-generated messages for extortion, while the militant group Boko Haram reportedly used AI tools to plan attacks and design weapons after circumventing existing safeguards.
Experts say the continent is operating in a safety vacuum. Jonathan Shock, an associate professor of mathematics at the University of Cape Town, told Rest of World that “there’s essentially no real discussion about AI safety.” He warned that without coordinated safeguards, African nations risk exposing populations to unchecked AI behaviours, and that fragmented national regulations could raise costs and complexity for U.S. and Chinese firms seeking market entry.
The ability to deactivate a model is emerging as a key risk factor. Closed-weight systems such as ChatGPT and Claude remain under direct control of their developers, whereas open-weight models like China’s DeepSeek and Kimi, as well as some American offerings, can be downloaded and altered by users, making remote shutdown difficult. Jane Munga, a fellow at the Carnegie Endowment for International Peace, emphasized that the critical question is not the country of origin but “who holds the off switch.”
Despite lagging behind other regions in overall AI uptake, more than 80 percent of the 85 large African firms surveyed by PricewaterhouseCoopers are already running AI pilots, though many lack formal safety oversight. Mélanie Keïta, chief executive of Nairobi-based financing firm Melanin Kapital, said “everyone is just trying to catch up” on deployment. Kenya stands out as the sole African participant in an international AI safety network that evaluates advanced models, but the network’s tests have focused on generic global threats rather than continent-specific vulnerabilities.
Stakeholders are calling for safety assessments tailored to African contexts. Joanna Wiaterek, co-founder of the Centre for AI Security and Access, noted that language-specific safeguards performed unevenly during recent tests. Research scholar Gathoni Ireri of the Ilina Program argued that impact should be measured against each country’s economic size, because an incident deemed minor elsewhere could be catastrophic for an African nation. Somalia’s state minister for foreign affairs Ali Mohamed Omar urged the development of local capacity to conduct third-party evaluations that reflect regional priorities.