He says companies that fully outsource their “thinking” to AI labs risk losing control of their future.
Microsoft CEO Satya Nadella has sharpened a warning he first raised earlier this month, telling businesses that depending entirely on a single AI provider could threaten their survival. Speaking on CNN’s Fareed Zakaria GPS on Sunday, Nadella said companies need to think carefully about what they hand over to AI model makers, including their data and prompts.
Keep Your Own Data, Build Your Own Models
Nadella argued that companies should retain the metadata generated every time they use an AI model, so they can eventually train their own systems rather than relying solely on outside labs. Without that kind of control, he suggested, a business risks losing its independence entirely by effectively outsourcing its decision-making to external AI providers.
Coding Tools Are a Particular Concern
Nadella singled out AI labs’ built-in coding assistants such as Anthropic’s Claude Code and OpenAI’s Codex as a specific risk area. He recommended that companies separate the “harness,” or coding interface, and their institutional memory from the underlying model itself, allowing them to switch providers freely without losing continuity.
A Self-Interested Message — But Not Without Merit
Notably, Microsoft is a major investor in both Anthropic and OpenAI, and its own cloud business sells the kind of multi-model infrastructure Nadella is promoting. Still, his concerns echo worries already circulating in the startup world, including from investor Jason Calacanis, who cautioned founders earlier this year about AI labs potentially studying and replicating their ideas.
Consumers Face a Different Calculus
When asked how to protect users’ data, Nadella compared it to a price for the consumer’s service and presented the concept as an advertising-based business model that has existed in the market for as long as technology has been functioning.
