Microsoft CEO Satya Nadella has issued a stark warning to companies using AI, urging them to be cautious about the data they share with AI models. Nadella's concern revolves around the potential for AI labs to act as Trojan horses, gaining access to sensitive business information and using it to their advantage.
In a blog post, Nadella highlights the double cost of using AI models. While companies pay for token usage, they also unknowingly reveal valuable proprietary knowledge, which can be used by the AI labs to enhance their models. This knowledge, Nadella argues, is akin to institutional know-how, something that competitors could never buy but are being handed over by enterprises.
The CEO emphasizes the irony of AI companies freely training on public data while restricting others from doing the same to their models. He suggests that if AI labs can scrape the internet to train their models, it's only fair that enterprises should have the right to study and 'distill' those models in return. This distillation process involves using a model's outputs to learn its workings and train new, often cheaper models.
Nadella's solution involves companies retaining ownership of their data, including prompts and feedback, and building their own proprietary learning environments on the cloud. He also advocates for 'orchestration layers' that allow companies to switch between AI models from different providers, ensuring they are not locked into a single model. This approach aligns with the growing trend of open-source models, which are being installed on-premises by large enterprises.
The shift towards open-source models is evident in the market. Companies like Vercel and OpenRouter are seeing a surge in traffic to open-source models, with open models accounting for a significant portion of their traffic. This trend is further supported by the Linux Foundation's Agentgateway project, which aims to accelerate AI agent adoption while maintaining security, observability, and governance.
Nadella's warning is particularly relevant given Microsoft's investments in both OpenAI and Anthropic. As the company encourages enterprises to be wary of proprietary models, the trend towards open-source AI is likely to continue. Nadella concludes by emphasizing the importance of companies retaining ownership of the intelligence they create, stating, 'In consuming intelligence, you are creating intelligence. And what you create should belong to you.'
