Microsoft CEO Satya Nadella says people are paying AI companies twice, suggests 5-point fix to problem

Microsoft CEO Satya Nadella has coined a new term – “Reverse Information Paradox,” referring to how companies today risk giving away their knowledge to AI labs by using their tools.

In an article on X, Satya Nadella referred to Nobel Prize-winning economist Kenneth Arrow’s “Information Paradox.” The Microsoft CEO said Arrow described a problem in which the seller risks giving away knowledge in order to sell it because its value is known only after the buyer has it.

According to Nadella, AI creates the opposite problem. “In the AI age, the buyer risks giving away knowledge, just in order to use what they bought,” he wrote. That is, to make the most out of an AI tool, you must feed it proper information, and prompts, something that the AI can then learn from and get better.

Before we delve deeper into Nadella’s idea of a reverse information paradox, do note that under his leadership, Microsoft was one of the early backers of OpenAI, and continues to have a stake in the AI startup. The Redmond giant has over 78 different Copilot AI tools available for users with a focus on enterprise use. Apart from that, Microsoft is also pushing for its own family of MAI models.

AI models are learning from you, says Satya Nadella

A key issue Satya Nadella pointed out was that every client or user was paying two times for using an AI tool – money and information. “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 Microsoft CEO explained.

You see, AI companies can learn from the prompts and data you use, while restricting distillation of their models, something Satya Nadella called “ironic.”

Nadella said the issue requires more than standard data protection because models learn from “exhaust” – the prompts you write, the tools agents use, and the corrections people make when a model is wrong. “Every correction is distilled into institutional know-how,” he wrote, describing it as the kind of knowledge a competitor could never buy and the kind that can leak “trace by trace, correction by correction, eval by eval.” “In consuming intelligence, you are creating intelligence. And what you create should belong to you,” he added.

Over time, he argued, the information asymmetry becomes increasingly skewed. The seller learns more and more about the buyer through usage, while the buyer learns very little about what the seller is learning in return. “That is what I think of as the Reverse Information Paradox,” he wrote.

Satya Nadella raised concerns over this singular flow of learning. He explained, “If learning flows in only one direction, economic value converges toward the owners of the learning infrastructure rather than the creators of the knowledge itself.”

How to fix Reverse Information Paradox?

To address this, Nadella said enterprises need a real “trust boundary” for their human capital and token capital to compound. According to the Microsoft CEO, this boundary should cover an organisation’s data, traces, evals, adapted weights and memory, and that nothing, including “intelligence exhaust,” should cross it without consent.

Quoting Palantir CEO Alex Karp, Nadella wrote, “What the technical customers want is control over their compute, their models, their data stack, and their alpha. They want to know they own the means of production, and it’s not being transferred to someone else.”

The Microsoft CEO set out five things businesses must do. Starting with control – Nadella wants every enterprise to have their own evaluation for AI models, as well as the retention of its data. He also insisted on building capability – via creating proprietary learning environments where models can be trained without exposing company information.

Apart from these, Satya Nadella pointed out the importance of choice – not relying on a single AI model. This in turn can also help in cost – by using different models depending on cost efficiency for any given task. And lastly, Nadella said that these four pointers together will compound your AI investment over time.

To sum it up, Satya Nadella reinstated the Reverse Information Paradox, by stating that what we need today is for a company to be “able to use a model without giving up the knowledge that makes it unique.”

Do note that open-weight models that exist today are a step in this direction. Anyone can download and run these models, such as DeepSeek V4 or Zai’s GLM 5.2, locally on their systems, and fine-tune them according to their needs. All of this can happen, without you having to give away your data to the AI company.