In a remarkable demonstration of artificial intelligence capabilities, British AI startup Mantic has outpaced human experts in an international forecasting competition, marking a significant milestone in predictive analytics.
The competition, known as the Metaculus forecasting cup, assembles experts and enthusiasts quarterly to predict the likelihood of future geopolitical events. These challenges range from forecasting military developments to judicial decisions, requiring participants to assess the probability of occurrences weeks or months ahead rather than simply guessing yes or no.
Mantic’s AI system not only competed but exceeded expectations by achieving over 80% of the average score of the top human predictors, a notably higher performance than initially anticipated. When the contest began in June 2025, participants predicted the leading AI bot would manage only around 40% of the top human average score, underscoring the AI’s unexpectedly strong showing.
According to Toby Shevlane, Mantic’s CEO, this milestone is “kind of mind-blowing,” highlighting the advanced capabilities of AI in tasks that require complex judgment and forecasting. The Metaculus platform itself has a notable track record, with users previously predicting significant events, such as the Russian invasion of Ukraine two weeks before the event and forecasting the overturn of Roe v. Wade with a 90% probability nearly two months in advance.
Forecasting is a critical tool widely used across government agencies. Nathan Manzotti, an AI and data analytics expert with experience in the U.S. Department of Defense and other agencies, emphasizes that forecasting serves as essential decision support. However, he notes a limitation: the reluctance of leadership to rely on data when it conflicts with intuition, an obstacle that even advanced AI cannot overcome.
The success of Mantic’s forecasting AI demonstrates the progress being made in AI-driven predictive analytics. It has implications beyond contests, potentially influencing how decisions are supported in governmental and commercial sectors worldwide. Although AI cannot replace human judgment fully, its ability to provide accurate probabilistic predictions presents a new informative dimension for strategic planning and risk assessment.
This achievement also aligns with broader advancements in the UK’s AI ecosystem, exemplified by other startups like Deep.Meta, which utilizes AI to improve industrial efficiency and environmental outcomes. Together, these developments illustrate the growing role of AI startups in driving technological innovation with tangible societal benefits.