Microsoft chairman and CEO Satya Nadella has flagged a new economic problem at the heart of the AI industry, arguing that using AI models forces companies to pay twice: once in money, and once in the proprietary knowledge they must feed into a model to make it useful. In a note on X, Nadella wrote that over time, this creates a widening information gap where the model provider learns more about a company's operations, decisions and corrections, while the company learns almost nothing about what the provider is doing with that knowledge. Nadella frames this as a mirror image of Kenneth Arrow's Information Paradox, noting that AI creates the opposite dynamic where the buyer risks giving away its knowledge. He traces this to "exhaust": the prompts employees write, how agents are used, and above all the corrections humans make when a model gets something wrong, which he argues leaks out gradually. Nadella also takes aim at model providers who rely on fair use of public data but impose restrictive terms on customers while reserving the right to learn from customer data themselves. To address the imbalance, Nadella lays out five principles, including that companies should build their own private evaluations and retain ownership of their own data, traces, feedback and institutional context, and build proprietary learning environments within their own systems.