Microsoft CEO Satya Nadella has doubled down on his warning that companies relying entirely on proprietary AI labs for their AI needs will not survive.

Speaking on CNN’s ‘Fareed Zakaria GPS,’ Nadella argued that businesses must retain control over their data, prompts, and metadata when using AI models, so that they can eventually train their own models or use open models.

He specifically advised against relying on built-in coding tools from AI labs, such as Claude Code or ChatGPT Codex, which he called ‘harnesses.

‘ Nadella advocates for an architecture where the harness, model, context, and memory are all separate, allowing companies to switch between models without losing control.

This advice comes despite Microsoft being an investor in both Anthropic and OpenAI.

Nadella’s warning addresses a broader concern that AI labs might eventually offer competing services based on data shared by their enterprise customers, similar to platform risk long discussed in the startup world.

He noted that this risk grows as enterprises adopt AI agents with access to internal company data.

However, Nadella made a clear distinction between businesses and consumers, stating that for individuals, sharing data is simply the price of using free services, analogous to the advertising business model.