African leaders called last week for a greater role in setting global standards for artificial intelligence.
Their statements at the United Nations came amid warnings from Silicon Valley executives about AI models being misused, and sometimes going rogue and taking unauthorized actions. OpenAI’s agents hacked government websites across the world in June. Models from Anthropic, Google, and Meta, too, hacked into other companies’ systems during tests.
“Africa … must be equal co-architects in determining the standards, ethics, and architectures of this technological era,” Lewis Garseedah Brown II, permanent representative of Liberia, said at a U.N. Security Council meeting on AI.
African companies and governments are racing to deploy AI systems built in the U.S. and China. But many nations lack the ability to independently test whether those systems are safe for their populations, experts told Rest of World.
“There’s essentially no real discussion about AI safety. So, there’s a real vacuum,” said Jonathan Shock, associate professor in math at the University of Cape Town.
If countries set up their own safeguards, it could complicate operations for American and Chinese companies in new markets. Fragmented regulations could make global deployments of models more costly and difficult.
Incidents of AI misuse and malfunction are surfacing across Africa. Kenya’s predictive machine learning algorithm for setting health insurance contributions has driven up costs for poor households. Scammers in West Africa have used AI-generated content to blackmail victims in extortion schemes. The militant group Boko Haram has also used AI to organize attacks and design weapons after bypassing safeguards.
“I think it’s going to take a potentially catastrophic incident for governments to really take [heed],” Shock said, which he dubbed his optimistic view. “The pessimistic view is we don’t stop it in time.”
Race to catch up
Africa currently lags behind other regions in AI adoption, though its companies are moving quickly. Out of 85 large companies surveyed by PricewaterhouseCoopers this year, more than 80% were running AI pilots. Many lack formal structures to oversee safety, according to the report.
“Everyone is just trying to catch up [on AI deployment],” Mélanie Keïta, CEO and founder of Nairobi-based financing company Melanin Kapital, told Rest of World.
Like many African companies, Melanin Kapital has weighed the choice between American and Chinese AI models, testing both internally. Both come with their own risks, Keïta said. “We might become a product of other foreign corporations that … own all our data as opposed to really being autonomous and free.”
Risk also correlates with how easily a model can be shut down if something goes wrong. ChatGPT and Claude are closed-weight models, centrally controlled by their providers.
Chinese models such as DeepSeek and Kimi, along with a few American models, are open-weight and can be downloaded and modified by users. That makes it harder for their developers to shut down remotely, Jane Munga, fellow at the Carnegie Endowment for International Peace, told Rest of World.
“The distinction that matters is not Chinese versus American; it is: Who holds the off switch?” she said.
Lack of guardrails
Governments across Africa are beginning to use AI in sensitive areas such as healthcare, education, and public services, even though fewer than half of African countries have AI policies or strategies, Shock said.
Kenya is the only African nation participating in an international AI safety network that is developing methods to evaluate advanced AI models.
Last year, the network tested AI agents for risks such as sensitive data leaks and fraud. But the exercise focused on common global risks rather than ones specific to Africa, Joanna Wiaterek, co-founder of the Centre for AI Security and Access, told Rest of World. It found agent safeguards worked unevenly across languages.
“Africa-centric” safety evaluations should judge harm by the size of each country’s economy, Gathoni Ireri, research scholar at the Nairobi-based Ilina Program, an AI safety research center, told Rest of World. An incident that might be considered minor in a Western country could irrevocably harm an African one, she said.
Companies currently carry out their own safety tests, but this needs to be replaced by third-party evaluations that assess risks specific to Africa, Ireri said.
“I think we also have to have our own ideas about what it means for a system to be safe within our own context,” she said.
African nations need to develop the capacity to conduct those evaluations, Ali Mohamed Omar, Somalia’s state minister for foreign affairs, said at the U.N.
“Africa cannot remain merely a market for systems developed elsewhere, a source of raw data, or a testing ground for a technology it did not help govern,” he said. “[Africa] must be granted a genuine representation and a voice in these multilateral processes.”
Leave a Reply