Startups Face Risks in AI Infrastructure Dependencies

Numerous startups are currently building their operations on AI models and services that they do not directly own. This extensive reliance on external components can introduce significant risks, particularly if provider terms, technologies, or legislation change. Analysts emphasize that companies must be deliberate about which parts of their technology stack they own versus rent.

A central issue arises when companies fail to fully grasp their dependence on cloud providers' hardware, data governance, and third-party APIs. When these external factors shift—due to legislative changes or a provider's own business strategies, for instance—startups may find themselves losing control over essential functions. Notably, the U.S. CLOUD Act and FISA Section 702 grant U.S. agencies broad authority to access data held by American cloud providers without a warrant.

JB Benjamin, CEO of Veldris, an AI infrastructure company, stresses a holistic approach. He states that AI adoption must consider not only the core inference and datasets but the entire stack. Connectivity and storage components, along with geopolitical factors, can represent significant attack surfaces that companies often overlook.

Businesses operating in sectors like healthcare, finance, or defense must pay close attention to these infrastructure decisions. While building everything in-house is not always necessary, identifying and managing the elements of the technology stack whose loss or alteration could jeopardize business continuity is essential. Failure to make these strategic considerations can lead to unexpected and costly consequences.