In the context of countries racing to build artificial intelligence (AI) capabilities, Mr. John Roese - Global Technology Director and AI Director of Dell Technologies, said that many places are having difficulty trying to control and adjust a technology that changes too quickly.
Speaking at Dell's annual event (taking place from May 18 to 21, 2026) in Las Vegas (USA), Mr. Roese emphasized that the AI strategy of countries today needs to aim for a long-term vision, instead of focusing on controlling each specific model or technology.
According to Mr. Roese, the current reality shows that AI changes too quickly, making many regulations obsolete as soon as they are implemented.
He cited an example of the original European AI regulations, built before Retrieval-Augmented Generation (RAG) technology appeared. Previously, AI models mainly learned from pre-trained data, forcing businesses to transfer sensitive data to AI providers such as OpenAI or Anthropic if they wanted to apply this technology.
But the appearance of RAG has completely changed the operating method. This technology allows the AI model to securely connect to the company's private data without having to transfer all data to a third party. According to Mr. Roese, this makes many assumptions in previous management policies quickly lose value.
In addition, the Dell leader also believes that countries should build an overall AI framework, focusing on infrastructure capacity and long-term management capabilities.
He highly appreciated the UK's approach when it does not rigidly regulate technology, but sets a goal to increase the country's computing capacity (data processing ability and operation of artificial intelligence models) 20 times in the AI era.
According to Mr. Roese, this is appropriate thinking because a country does not need to know exactly what future technology will be like, but needs to prepare sufficient infrastructure to adapt to all changes.
He also proposed the "agent-AI" model, in which management agencies play the role of providing digital identity for operating AI systems. This approach helps management agencies to monitor, control and even disable the system when necessary.
However, Mr. Roese believes that the biggest challenge today is the lack of unity in global AI policy. For example, Dell is operating in more than 1,000 different jurisdictions with separate management requirements.
Roese warned that complying with too many complex regulations could sharply increase the cost of AI deployment. He called on policymakers to carefully calculate the economic impact before issuing new regulations.
In terms of technology, Roese supports the "hybrid AI" model, combining on-site infrastructure and cloud computing. According to him, no large enterprise actually puts the entire system on a single cloud platform because cost, security and flexibility are always complex problems.
Mr. Roese believes that the "hybrid AI" model will be especially suitable for countries that want to build independent AI technology sovereignty and infrastructure.
