Sovereign AI push widens as SoftBank, Sony, NEC, Honda set up Noetra

The article traces the idea to a familiar business problem: core systems built on one vendor's technology, long known as vendor lock-in. It argues the risk is larger for AI because outputs and training data are harder to inspect than in conventional IT, and because AI is being woven into more of the work a company depends on.

The piece lays out the familiar complaints behind sovereignty: questions, orders and sensitive corporate, government, financial and medical data processed on overseas cloud servers, subject to local law; AI services whose fees, features or very existence can change quickly; and models trained on global data that may not match Japanese usage, business custom and legal or medical systems. The article also notes one country's government has ordered a specific AI service halted over security risk, and warns that relying on a single vendor — even a domestic one — does not preserve sovereignty either.

Its survey of the field contrasts the US stance with the EU's early 'digital sovereignty' push and its AI Factory data centers, China's state-led drive for self-reliance spanning models, chips, cloud and data, and India's push built on its large digital population and many languages. Japan, the article says, is choosing differently. Rather than chase giant general models, it is aiming at manufacturing, robots, healthcare, finance and public administration, and at small models for edge devices such as robots and factory equipment. Whether that bet pays off likely hinges on the five-layer ecosystem the article describes — models, data, data centers, hardware components, and people and operations — being built out evenly rather than in patches.