An essay argues that enterprises risk revealing proprietary knowledge when using AI, because models learn from prompts, workflows, evaluations, and corrections, creating institutional know-how.
Microsoft CEO Satya Nadella warned that enterprises risk giving away proprietary knowledge every time they use artificial intelligence.
In an essay titled "The Reverse Information Paradox" Nadella argued businesses pay for intelligence twice: once with money and again by revealing the proprietary knowledge required to make models useful.
It is worth remembering Microsoft holds a roughly 27% stake in OpenAI and has integrated those models into Azure AI, Microsoft 365 Copilot and GitHub Copilot.
"In consuming intelligence, you are creating intelligence. And what you create should belong to you," Nadella wrote.
He said protecting enterprise knowledge requires more than securing raw data because models also learn from prompts, workflows, evaluations and the corrections users make over time, and those interactions accumulate into institutional know‑how rivals cannot easily replicate.
London School of Economics professor, Luis Garicano, described the essay as smart economic thinking and Microsoft AI executive Nicolas Bustamante framed the issue as enterprises increasingly accumulating learning, not just data, and seeking to own the intelligence produced through their interactions.
In June Nadella had warned that an AI landscape dominated by a few models could concentrate economic value and weaken companies' competitive advantages, and he has called for a broader AI ecosystem.
