Responsible AI
Foundation-Model Concentration Risk: A New Item for the Board Risk Register

Jeremiah Ssekabira
June 1, 2025 · 8 min read
Executive Summary
Boards underwriting AI strategies should pay close attention to a quiet but consequential exposure: concentration on a small number of foundation-model providers. This article explains why concentration risk is now a board issue and what governance responses are available short of vertical integration.
Context
Most enterprise AI today is built on a small set of foundation models. Those models are improving rapidly, but their providers also change prices, terms of service, content policies and model availability with limited notice. Organizations that have rebuilt customer-facing experiences around a single model can find their service quality, cost base and policy posture shifting on someone else's schedule.
Key Issues
- Single-provider lock-in across multiple business processes.
- Limited transparency into training data, evaluation methods and incident history.
- Geopolitical and export-control exposure on foundation-model access.
- Cost trajectories that can change materially between contract cycles.
Strategic Implications
Concentration risk is a board-level risk because the failure mode is enterprise-level. A unilateral pricing change or content-policy shift can disrupt revenue, service and compliance simultaneously.
Governance Considerations
- Treat foundation-model providers as critical suppliers with corresponding due diligence.
- Require an abstraction layer between application code and any single model provider where feasible.
- Maintain at least one tested alternative provider for high-impact use cases.
Practical Recommendations
- Add AI supplier concentration to the enterprise risk register and report it to the board annually.
- Negotiate transparency, evaluation and incident-notification clauses into vendor contracts.
- Stress-test the organization's AI dependency by simulating a 30-day loss of the primary provider.
Conclusion
Concentration risk is governable, but only if the board names it. Treating foundation-model providers as commodity SaaS underestimates the strategic dependency. Naming the exposure is the first step to managing it.
Share this article
Bring AI Governance to Your Boardroom
Book an executive masterclass or advisory session with Jeremiah Ssekabira.
Previous
Re-engineering the Three Lines of Defense for the AI Era
Next
AI in the Public Sector: A Governance Posture for African Institutions
The AI Governance Brief
Weekly insights on AI governance, board leadership, responsible AI and emerging technologies — delivered to executives across Africa and beyond.