Satya Nadella, CEO of Microsoft, has raised concerns that proprietary AI models could capture sensitive corporate knowledge and evolve into competitors. His warning carries significant weight given Microsoft's investments in OpenAI and Anthropic—both companies increasingly gaining access to business data from enterprises using their tools.
Nadella argues that AI users pay twice: once for tokens and again through their data, which becomes part of training these systems. He contends that if AI companies can freely train on internet-wide data, businesses should have reciprocal rights to study or distill those models for their own use.
The tension reflects growing corporate caution around proprietary AI solutions. According to TechCrunch, one infrastructure startup reports a significant shift toward enterprises installing open-source AI models on-premise, signaling what it calls the next major wave in corporate AI adoption. For companies concerned about data leakage and vendor lock-in, this trend suggests orchestration layers that switch between different providers' models could become standard practice.
