Microsoft CEO Satya Nadella published a blog post on Monday that has sent a jolt through the enterprise AI world, warning that companies using proprietary AI models may be unknowingly handing over their most sensitive business information to the very companies that could become their competitors.
Nadella joins a growing chorus of tech leaders — including venture capitalist Jason Calacanis and Palantir CEO Alex Karp — who have raised alarms about the Trojan horse nature of proprietary AI systems. The concern is straightforward: startups and enterprises feed these models with proprietary knowledge to make them useful, and in doing so, train the models on the very insights that give them a competitive edge.
Also read: Taiwan's AI Cycle and 2H Outlook: DBS Weighs In
Nadella’s core argument: You’re paying twice
In his post, Nadella argued that AI users — whom he called the “buyers” — are paying for intelligence twice. “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,” he wrote. “The better you want the model to perform, the more of that knowledge you have to feed it!”
Nadella emphasized that enterprises are literally teaching the models about their business nuances through what he called “exhaust” — the prompts users 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 wrote, calling it “the kind of knowledge a competitor could never buy.”
Also read: OpenAI hires family-focused product manager as ChatGPT user base ages
The distillation hypocrisy
A central point in Nadella’s argument concerns the practice of “distillation” — using a model’s own outputs to train a new, often cheaper, model. In February, Anthropic accused Chinese open-source models of sending millions of prompts to Claude to improve their own systems, urging U.S. government action on export controls.
Nadella called out what he sees as hypocrisy: “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.” He specifically warned about model makers that “reserve the right to learn from customer usage and interaction data.”
Nadella’s solution: Build your own AI environment
The Microsoft CEO’s proposed solution aligns closely with his company’s cloud business. He urged enterprises to “retain ownership” of their data — including prompts and feedback — and build “proprietary learning environments” on the cloud, which could mean Microsoft’s Azure. He also recommended creating “orchestration layers” that allow easy switching between AI models from different providers, avoiding lock-in to any single vendor.
Tools like AI gateways that enable this switching have become increasingly popular. While Nadella never explicitly mentioned open-source models, the subtext is clear: retaining ownership often means running models on your own infrastructure.
The industry is already moving
Idit Levine, founder and CEO of Solo.io — which makes networking and security software for enterprise AI systems — told TechCrunch she’s seeing this shift play out with her own customers. After experimenting with proprietary models, they start asking: “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,” she said. “They understand that, and they can control it.”
Solo.io’s technology was selected last year as the tech powering the Linux Foundation’s Agentgateway project. Her company counts enterprises like T-Mobile, ADP and SAP as customers. She sees on-premise open-source models as the next big wave in enterprise AI use.
Other companies are reporting similar trends. Vercel, best known as a platform for building and hosting websites, recently added AI model-switching tools. OpenRouter, which helps developers route requests across different AI models, is also seeing a surge in traffic to open-source models. Open models accounted for 29% of all traffic routed through Vercel’s gateway last month.
With the CEO of Microsoft — a company that has invested billions in both OpenAI and Anthropic — now openly urging enterprises to be wary of proprietary models, the momentum toward open-source, self-hosted AI solutions appears likely to accelerate. As Nadella wrote: “In consuming intelligence, you are creating intelligence. And what you create should belong to you.”
Frequently Asked Questions
What did Satya Nadella say about AI data risks?
Nadella warned that enterprises using proprietary AI models are paying twice: once for token usage, and again by handing over valuable proprietary knowledge that the model makers could use to compete against them.
What is AI model distillation and why is it controversial?
Distillation is the practice of using a model’s own outputs to train a new, cheaper model. Nadella criticized model makers for restricting distillation while freely training on public data, calling it hypocritical.
What solution did Nadella propose for enterprises?
He urged companies to retain ownership of their data, build proprietary learning environments on the cloud, and use orchestration layers to easily switch between AI models instead of being locked into one provider.
How are enterprises responding to these AI data concerns?
Many large companies are moving toward open-source AI models installed on their own premises, which provide similar performance at lower cost while giving them full control over their data.
