Exploring Sovereign AI: Balancing Control, Resilience, and Investment

During the RAISE summit in Paris, an unexpected power outage plunged the event into darkness at a critical moment when Arthur Mensch from Mistral and Mark Surman from Mozilla were discussing the importance of open-source and sovereign AI. As the microphones fell silent and screens went blank, their discussion on resilience and the necessity of having multiple options in AI became vividly apparent.

The power failure underscored a key point in their dialogue: the significance of having alternatives when a supplier or platform fails. In this context, open-weight and open-source models offer users more control and adaptability, essential features that businesses and governments are increasingly seeking.

The trade show floor of RAISE was abuzz with the term “sovereign AI,” indicating a strong desire from corporations and governments to maintain control over the AI technologies they depend on. This desire is not only about power but also about mitigating risks such as loss of oversight and dependence on opaque AI systems, which companies and investors frequently discuss.

Investment Opportunities in Responsible AI

Investors are already eyeing opportunities within what is being called the responsible AI stack. According to a global survey conducted by ReframeVenture and ImpactVC, nine out of ten venture investors perceive financial potential in this field. However, identifying these opportunities remains a challenge as numerous international bodies like the UN, OECD, and UNESCO have outlined various principles for "good AI", focusing on transparency, accountability, and privacy.

The task now is to translate these principles into actionable market strategies, prompting questions on who will demand these AI features and where companies can find value.

Agency and AI Stack Optionality

Agency in AI isn't just about having a “human in the loop.” It involves having practical options across the AI stack, including data, infrastructure, models, and applications. This concept translates into who controls data rights, how adaptable the infrastructure is, and the degree to which applications allow for user intervention and control.

Agency is about the level of control and visibility individuals or enterprises maintain, not just a single feature. This control is crucial as it can impact everything from product margins to user engagement.

Risks of Dependence in AI

The risks associated with dependency in AI are becoming increasingly evident. Companies relying on the same AI models and infrastructure as their competitors face significant vulnerabilities. For example, changes in API pricing can critically impact margins, and access issues can disrupt services, as seen with Anthropic’s Mythos release.

For investors, the critical question is whether a company has control over its AI stack or is merely dependent on another provider's technology. This control affects everything from market positioning to the potential for enduring value creation.

Organizations desire AI solutions that enhance productivity without compromising proprietary data or customer relationships, while governments and developers look for flexibility and independence from major foreign providers.

Investment Committees: Key Questions

Investment committees should focus on how agency influences investment quality and potential value. Five essential questions should guide their evaluations:

1. Where does agency fit within the investment thesis? Evaluate whether AI-driven companies maintain control over critical data and models, and if this control can differentiate their products.
2. Is agency accounted for in the valuation? Determine whether the company’s value accounts for stronger growth, market access, and retention due to agency.
3. Who is willing to pay for agency now, and why? Identify the customers who value agency and understand the problems it solves for them.
4. Can agency lead to lasting value capture? Assess whether control over AI systems can help win new customers or deepen market integration.
5. Is agency becoming a vital control point? Determine whether the company is solving essential feature issues or creating a critical layer within the AI stack.

Balancing Returns and Responsibilities

While agency enhances resilience and customer satisfaction, it also aligns with broader social responsibilities. The UN's Independent International Scientific Panel on AI has emphasized the importance of using AI to complement human skills and distribute benefits widely.

Although agency cannot ensure such outcomes, it provides the framework to adapt AI deployment in ways that align with public and institutional mandates. For example, pension funds and sovereign wealth funds are encouraged to consider the long-term economic implications of AI on productivity and market stability.

The convergence of agency-related opportunities is clear as AI authority expands. Investors are increasingly focused on maintaining control over the components of AI systems, a shift reflected in the strategies adopted by specialized funds and traditional asset allocators.

Ultimately, the key question for investors is not just how much agency machines will acquire, but how much people and organizations are willing to pay to retain their own.