Artificial Intelligence to transform safety in Kenya's mining sector

As Kenya prepares to expand its mining sector, from gold operations in Kakamega to emerging opportunities in critical minerals at Mrima Hill in Kwale County, concerns around safety and sustainability have taken on renewed urgency.

Last week, the American Chamber of Commerce–Kenya (AmCham Kenya), together with the US Embassy in Kenya, held a one-day US–Kenya Critical Supply Chains Conference in Nairobi, where Cabinet Secretary for Mining, Blue Economy and Maritime Affairs Hassan Joho said Kenya is ready and open for investment in the mining sector.

However, recent fatal incidents at mining sites across the country have brought safety concerns into sharp focus. In December last year, three male artisanal miners died after a gold mine collapsed in Savane, Kakamega County. In June, four others were killed when a mine caved in at Karon, Turkwel, in West Pokot.

According to police, dozens of people have lost their lives in mining accidents in recent years, with dangerous sites spread across Siaya, Migori, West Pokot, Moyale, Kakamega, Nandi and Transmara counties.

Amid these developments, the sector faces a defining question: how can Kenya better protect artisanal miners working in complex, risky and dynamic mining environments? Khadija Said, a Kenyan mine safety expert at Pennsylvania State University in the United States who is pursuing a PhD in Mineral Engineering, says a crucial part of the solution lies in artificial intelligence (AI).

She argues that AI can be used to engineer intelligent systems that predict and control hazards before disaster strikes, rather than reacting after lives are lost. “For decades, mining hazards have been treated as sudden and unavoidable acts of nature. But hazards are  quantifiable physical processes that evolve over time and can be monitored and controlled with the right tools,” says Said.

She explains that AI sits at the centre of modern safety systems, supported by empirical and engineering-based methods. This approach is demonstrated in her peer-reviewed study, An artificial intelligence-based model for the prediction of spontaneous combustion liability of coal based on its proximate analysis, published in Combustion Science and Technology in 2021.

Said says the same approach can be applied to other risks, including mine collapses, radiation exposure and battery fires. “Most hazards evolve gradually, marked by changes in physical parameters. By monitoring these, AI models can predict unsafe conditions early, creating time for intervention,” she explains.

“These models rely on sensors that collect real-time data and feed it into controllers, essentially small computers, to predict danger,” she adds.

Contrary to popular belief, AI-based systems are not limited to large-scale mines. “They are even cheaper for artisanal operations, which use smaller pits and simpler systems,” she says, noting that most fatal accidents occur in such settings.

Unlike conventional alarm-based systems, AI models identify dangerous patterns across multiple variables, allowing early detection of unsafe conditions. “If you only respond to alarms, you are already late. Prediction allows us to act before failure occurs,” Said warns.As Kenya explores critical minerals such as niobium, often associated with uranium and thorium-bearing formations, predicting radiation exposure becomes vital. Ms Said says predictive models can enable proactive risk management through improved ventilation design and operational planning, rather than reactive shutdowns.

Similarly, monitoring stress accumulation and deformation in rock structures can allow engineers to predict mine collapses before they happen.

With mining operations increasingly shifting to electric vehicles, lithium-ion batteries present another growing safety risk due to thermal runaway, a dangerous process that can lead to fires and explosions. “In my current work, I treat thermal runaway not as a sudden event, but as a process,” she says. “It begins with internal heat build-up, followed by material breakdown, gas release and intense heat emission,” she continues.

By modelling temperature and gas changes, Ms Said has developed predictive tools that can identify unsafe conditions before battery failure occurs. Her findings have been presented at major international mining conferences, including the North American Mine Ventilation Symposium and the Society for Mining, Metallurgy and Exploration (SME).

Sense and predict risk

Her current research focuses on developing self-regulating control algorithms. “The goal is to design mines that can sense, predict and control risk,” she says. “When AI is grounded in physics and engineering, safety becomes something we plan for, not something we react to when it is too late.”

As Kenya positions itself within global supply chains for critical minerals, Said argues that safety standards will ultimately determine the country’s competitiveness.

However, she blames the persistence of fatal mining accidents on what she calls a major disconnect between research institutions and industry.

“Many mining companies do not deploy modern safety systems, even though they are safer and more efficient,” she says.

Drawing comparisons with South Africa, where she previously conducted research, Ms Said notes that the mining industry there actively partners with universities and funds applied research to solve operational challenges. Some of the AI-based models she helped develop are already in use in Mpumalanga coalfields. “Our people, both those working in mines and those living nearby, deserve better,” she says. “The government must enforce stronger safety systems across all mining operations.”