AI’s biggest risk is becoming a geopolitical dependency

Artificial intelligence (AI) poses a significant threat by becoming intertwined with geopolitical dependencies. The infrastructure that underpins AI technologies is increasingly being viewed as a strategic asset, prompting governments to implement export controls and other measures to restrict access. Companies often overlook the importance of considering who controls the AI models they choose, leading to reliance on foreign platforms that can be abruptly cut off by overseas governments.

Anthropic, for example, was forced to suspend its most advanced AI models by the U.S. government, highlighting the rapidity with which access can be revoked. This risk is not hypothetical; it has already led to reviews by governments and increased scrutiny on the economic and security implications of lost access to frontier AI.

Governments and companies are now prioritizing "sovereign AI" – control over data centers, chips, models, and applications within their own borders. However, many countries lack credible home-grown alternatives, with nearly 90% of the world's AI compute, investment, and leading foundation models concentrated in the U.S. and China.

This concentration leaves businesses globally dependent on foreign AI, particularly in Europe where over 80% of digital infrastructure and services are supplied by overseas companies, predominantly American.

Geopolitical tensions, such as tariffs, export controls, and the US Cloud Act, further complicate matters by governing the laws that apply to the storage and access of corporate data. This uncertainty forces companies to reconsider their AI procurement strategies. Instead of investing in the latest frontier models, companies should prioritize solutions that directly enhance productivity.

They should also maintain flexibility within their AI stacks, diversify their providers, and avoid being locked into a single vendor or regulatory framework. Companies should conduct tests to assess the impact of a sudden vendor disappearance and ensure that critical business functions can continue with minimal disruption.

Companies that build optionality into their AI strategies, akin to diversifying supply chains and avoiding single points of failure in cybersecurity, will be better positioned to navigate the geopolitical risks associated with AI. A recent Gartner report predicts that one in five cloud workloads will shift from global to local providers by 2027 due to geopolitical risks.

Airbus, for instance, has begun migrating critical applications to a French cloud provider while retaining the use of U.S. platforms for less sensitive workloads. This strategic diversification is a prudent approach as companies seek to secure their most sensitive data and infrastructure, ensuring they remain resilient against sudden reversals in AI geopolitics. Ultimately, policy certainty should be considered as crucial as access to AI technology itself.