Microsoft CEO Satya Nadella has issued a stark warning to businesses using artificial intelligence (AI) models from proprietary labs like OpenAI and Anthropic. According to Nadella, companies are essentially paying twice: once with money for AI token usage, and again by inadvertently sharing valuable business information.

The concern is that as startups and enterprises use these models, the labs gain insight into sensitive company data. This could potentially allow model makers to become competitors to their own customers, argues Nadella.

Nadella highlights a particular risk: enterprises are teaching AI models about the intricacies of their businesses through corrections and feedback. These corrections contain institutional knowledge that rivals could never buy on the open market, yet companies are handing it over without realizing the extent of the exposure.

In response, Nadella suggests that companies should retain ownership of their data by building proprietary learning environments in the cloud. He advocates for creating orchestration layers to easily switch between AI models from different providers, allowing enterprises to avoid being locked into a single service.

The trend towards using open-source AI models on-premises is gaining momentum. Companies like T-Mobile and ADP are increasingly installing these models because they offer significant cost savings while still delivering 90% of the functionality of proprietary alternatives.