Nigerian AI researcher and entrepreneur Sanmi Koyejo is playing an increasingly prominent role in efforts to make artificial intelligence more reliable, accurate and secure.
Koyejo, co-founder of AI security startup Virtue AI, has spent much of his career examining what AI models can and cannot do, with a focus on ensuring that increasingly powerful systems perform reliably when deployed in the real world.
His work has now attracted the attention of Meta, which in June announced that it had hired Koyejo and Virtue AI co-founders Dawn Song and Bo Li to join its Superintelligence Labs, where they will contribute to the company’s efforts in AI security and research.
Koyejo is expected to play an important role in Meta’s Fundamental AI Research (FAIR) division, one of the company’s key research arms focused on advancing artificial intelligence.
His move to Meta comes after years of research into one of the industry’s biggest challenges: determining whether AI systems can actually be trusted to perform as expected outside controlled testing environments.
Koyejo is also the director of Stanford University’s Trustworthy AI Research Lab (STAIR), where he oversees research into AI measurement science.
The field focuses on pressure-testing AI models and developing better ways to measure their performance in real-world settings.
Before AI systems are released to the public, they typically undergo a series of evaluations and benchmark tests. But Koyejo argues that success in a laboratory or on a particular benchmark does not necessarily mean a model will perform equally well when used by hospitals, schools, businesses or government agencies.
The difference between benchmark performance and real-world performance can be substantial.
For Koyejo, AI models should therefore be subjected to standards capable of demonstrating how reliable they are in the environments where they will actually be used.
He has compared this approach with established practices in fields such as medicine and auditing, where performance and reliability are tested against real-world outcomes.
The goal, he argues, is to give organisations a clearer understanding of what an AI evaluation actually means for their particular circumstances, including the level of uncertainty associated with the results.
That work has become increasingly important as AI systems are being deployed across sensitive areas, from healthcare and education to finance, government and cybersecurity.
Koyejo’s career has also included significant work to promote greater representation in the artificial intelligence industry. He served as president of Black in AI, an organisation focused on increasing the participation of Black researchers in the field, for five years.
Through his academic research, entrepreneurship and work with the wider AI community, Koyejo has built a career around a question that is becoming increasingly important as artificial intelligence advances: not simply what AI can do, but whether people can trust it to do those things safely and consistently.
His appointment to Meta’s Superintelligence Labs places the Nigerian AI expert at the centre of that conversation, as technology companies race to develop increasingly capable AI systems while grappling with the risks that come with them.
Vanguard News
The post Meet Sanmi Koyejo, Nigerian professor named in 2026 TIME100 AI List appeared first on Vanguard News.