Artificial intelligence users should retain ownership of the business knowledge generated through their interactions with AI systems, Microsoft Chief Executive Satya Nadella said, arguing that enterprises risk giving away valuable intellectual assets when relying heavily on proprietary AI models.
In a blog post published on Sunday, Nadella said companies effectively pay twice for AI services: first through usage fees and again by exposing proprietary information that helps improve the underlying models.
Enterprises Urged to Protect Institutional Knowledge
According to Nadella, businesses provide increasingly valuable operational knowledge as they refine AI outputs through prompts, feedback, and corrections. He argued that this information can become a strategic asset for model developers if customers do not maintain control over it.
“You essentially pay for intelligence twice, once with money, and again with something even more valuable: the proprietary knowledge you must reveal to make that intelligence useful. The better you want the model to perform, the more of that knowledge you have to feed it!”
Nadella said AI systems continuously learn from user interactions, including prompts, agent workflows, and corrections made when responses are inaccurate.
“Models learn from ‘exhaust,’ the prompts people write, the tools agents use, and especially the corrections people make when the model is wrong. Every correction is distilled into institutional know-how.”
He described that accumulated expertise as information that competitors could not easily acquire through conventional means.
The Microsoft chief also questioned restrictions imposed by some AI developers on model distillation—the practice of training a smaller model using the outputs of a larger one. Nadella argued that if AI companies benefit from training on publicly available internet data, enterprises should similarly have the ability to learn from the models they use.
“While the great innovation that comes from model providers having fair use rights to train models on public data is needed, I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation.”
Call for Flexible AI Infrastructure
To address those concerns, Nadella encouraged organizations to retain ownership of prompts, feedback, and other AI-generated data by building proprietary learning environments within their cloud infrastructure. He also advocated using orchestration layers that enable businesses to switch between different AI models instead of depending on a single provider.
His comments come as some enterprises increasingly evaluate open-source AI models that can be deployed within their own data centers or private infrastructure, offering greater control over data and operating costs.
Idit Levine, founder and chief executive of enterprise software company Solo.io, said many customers initially experiment with proprietary AI services before considering self-hosted alternatives.
“Can I take an open source model and run it on-prem? It will do almost 90% of what the big one’s doing. It will cost way less,” Levine said. “They understand that, and they can control it.”
Levine said she expects adoption of on-premises open-source AI models to continue growing among enterprise customers seeking greater governance over their AI deployments.
The trend is also being reflected by companies that provide AI routing services. Vercel reported that open-source models accounted for 29% of the traffic handled through its AI gateway last month, while OpenRouter has also seen rising demand for open models.
Although Microsoft has invested heavily in AI developers including OpenAI and Anthropic, Nadella’s remarks suggest enterprises should carefully consider how they manage the knowledge generated through AI usage.
“In consuming intelligence, you are creating intelligence. And what you create should belong to you,” Nadella wrote.