On Sunday, Microsoft CEO Satya Nadella issued a stark warning to businesses relying heavily on artificial intelligence (AI) models. He stated that firms which do not retain control over their own AI training data may not survive in the long run.

Nadella argues that companies should keep their metadata and trained parameters separate from proprietary AI labs, allowing them to develop internal AI solutions down the line. He warns against outsourcing thinking entirely as there is a risk of rival services emerging from the same providers who could potentially copy and outcompete clients.

The move comes amid Microsoft’s investment in major AI players Anthropic and OpenAI, hinting at self-serving interests alongside broader strategic concerns about data sovereignty and control. Enterprises are increasingly recognizing the need for multiple model options, especially those they can fine-tune independently. However, this shift could lead to a complex landscape of managing various coding agents from different providers.

While Nadella’s warning is primarily directed at businesses, his cautionary tone may resonate with broader debates around data privacy and technological dependence in the AI era.