July 14 – Microsoft CEO Satya Nadella has cautioned businesses against becoming overly dependent on proprietary artificial intelligence models, arguing that companies may be giving away valuable intellectual property every time they interact with AI systems.
In a blog post published on Monday, Nadella said organizations effectively pay twice when using AI services: first through subscription or usage fees, and again by sharing proprietary business knowledge that helps AI models generate more accurate responses.
According to Nadella, the more context, documents, and corrections companies provide to AI systems, the more they expose unique operational knowledge that could become valuable to model providers.
He explained that AI models continuously improve from user interactions, including prompts, tool usage, and corrections made when responses are inaccurate. Over time, these interactions can capture insights into how a business operates—knowledge that competitors would otherwise struggle to obtain.
Nadella argued that companies should retain ownership of this information rather than allowing it to become part of a model provider’s learning process.
More Read
Calls for Greater Control Over Enterprise AI
The Microsoft chief also questioned restrictions imposed by some AI companies on model distillation—the process of using an AI model’s outputs to train another model.
He suggested there is an inconsistency in allowing AI developers to train models using publicly available internet data while preventing customers from applying similar techniques to models they pay to use.
Nadella said enterprises should be able to preserve ownership of their prompts, feedback, workflows, and other data generated while using AI systems.
Push for Open and Flexible AI Infrastructure
To reduce dependence on a single AI provider, Nadella encouraged businesses to build proprietary learning environments and adopt AI orchestration layers that allow them to switch easily between different AI models instead of becoming locked into one platform.
Although he did not explicitly advocate for open-source AI, industry observers say his comments strongly support greater adoption of open models that organizations can deploy within their own infrastructure.
Many large enterprises are already moving in that direction by running open-source AI models on their own servers, allowing them to maintain tighter control over sensitive business data while reducing operating costs.
Industry executives report growing interest in self-hosted AI solutions, with organizations finding that modern open-source models can deliver performance close to leading proprietary systems at a significantly lower cost.
Companies providing AI infrastructure and model-routing services have also reported increased adoption of open-source models as businesses seek greater flexibility, lower expenses, and stronger data privacy.
Nadella concluded that organizations should recognize the value of the intelligence they generate while using AI, arguing that the knowledge created through those interactions should remain under the control of the businesses that produce it rather than becoming an asset for AI providers.
