Microsoft: "Sovereign AI Isn't Domestic AI... We Need to Secure Choice Across the Ecosystem" [Global AI Briefing]

[Financial News] As the race for AI supremacy accelerates, experts have pointed out that AI sovereignty (Sovereign AI) does not simply mean developing domestic AI models. The true meaning of Sovereign AI lies not only in developing specific AI models directly, but also in who controls the entire ecosystem—including data, AI models and computing infrastructure—and whether those options can be chosen when needed.
On the 5th (local time), Microsoft explained the concept of and technologies related to 'AI sovereignty (Sovereign AI),' which it announced with NVIDIA, in a cloud blog post.
In a report, Didier Ongena, Microsoft's public sector and sovereign lead, emphasized, "Sovereign AI encompasses not only where data is stored, but also which AI models process the data, who manages the infrastructure that powers AI, and whether AI can continue to operate when disconnected from external environments."

The report particularly emphasized that the ability to choose across the entire AI ecosystem is central to Sovereign AI.
He noted, "Companies and governments must be able to choose models suited to their tasks rather than becoming dependent on a single AI model, and switch to another model when a better one becomes available." The assessment is that excessive dependence on a particular model or cloud makes it difficult to respond to technological change as AI technology advances rapidly.
He also stressed that where AI is operated is important. Although routine tasks may be handled in a public cloud, sensitive information held by financial and public institutions must be processable in their own data centers or in a Private Cloud.
He added that in places such as manufacturing sites, where connecting to external networks is difficult, AI may also need to be operated directly on-site.
Against this backdrop, voices are growing in South Korea's industry that the country—which is concentrating its policy efforts on securing its own foundation models—must obtain control over the entire ecosystem, including data, GPUs, data centers, cloud services, security, model choice and operational capabilities, rather than focusing solely on securing models, to achieve AI sovereignty.
An industry official said, "The government and industry must recognize that the dependence of domestic AI development on the cloud and platforms of certain global companies has actually been increasing recently."
[email protected] Lee Gu-sun Reporter