Boards Urged to Demand Evidence-Based AI Governance Amid Rising Incidents

**TL;DR** Boards must demand verifiable evidence for AI system performance and risk. This shifts focus from innovation-led deployment to evidence-led governance, critical as AI-related incidents rise despite widespread adoption.

Artificial Intelligence (AI) adoption rapidly expands across industries, promising significant growth and efficiency. Yet, this expansion also introduces governance and risk management challenges. Boards face increasing calls to ensure AI systems are not only innovative but also demonstrably reliable and safe.

The 2025 Stanford AI Index reveals a concerning trend: AI-related incidents are increasing, even as global AI adoption accelerates. This indicates a growing gap between deployment speed and risk mitigation.

Many organizations actively use AI, but robust governance and risk management practices remain uncommon. McKinsey research indicates that while most companies deploy AI in some form, few embed strong oversight.

A fundamental principle underscores this emerging challenge: "Performance cannot be trusted without evidence." This emphasizes that claims of AI capability require verifiable data and testing, not mere assumption.

The absence of rigorous proof for AI system behavior creates significant organizational exposure. Boards, responsible for outcomes and accountability, must demand clear evidence regarding an AI system's defined purpose, scope, and acceptable risk thresholds.

This requires comprehensive testing for accuracy, bias, and real-world impact, moving beyond controlled environments to operational conditions. Documented validation processes are essential before and during deployment.

Continuous monitoring and transparent traceability of AI decisions are also critical. Systems evolve, and data shifts post-deployment, demanding ongoing oversight to ensure risks are managed and outcomes are explainable.

For countries like Nigeria, where AI offers pathways to accelerated growth and improved service delivery, the stakes are particularly high. The financial, regulatory, and reputational costs of AI failure are substantial, demanding a narrower margin for error.

Boards must watch next for the consistent implementation of these evidence-based governance frameworks, proving AI systems work as intended and risks are effectively managed.