Companies Could Go Bankrupt if Using AI, Microsoft CEO Warns
Microsoft CEO Satya Nadella has issued a serious warning to companies that rely too heavily on Artificial Intelligence (AI) to run their operations. According to him, a company faces the potential for collapse if it shares too much sensitive information with AI models.
He emphasised that business operators should be more vigilant regarding the data and instructions (prompts) they share with AI systems. “In my opinion, any company that does not have full control over its data will not last long. Because, essentially, you have outsourced your company’s thinking process,” Nadella stated, as reported by Futurism.
He recommended that companies should store all data related to how their employees use AI tools. By doing so, companies can eventually train or refine their own open-weight AI models. Furthermore, Nadella urged companies to begin moving away from closed-source agentic AI tools. Instead, he suggested using specialised AI platforms that can integrate various types of AI models, such as more affordable, open-source platforms.
“By separating the framework from the model, and separating the context and memory from that model, you can use various AI models according to their respective strengths,” he explained. “At the same time, any model can be replaced or removed, so that you maintain control over the direction and future of your own business,” Nadella added.
Although Microsoft is at the forefront of the AI development race alongside many other tech giants, Nadella has been proactive in voicing concerns regarding the technology’s potential negative impacts. Previously, he warned that AI users might unconsciously “pay twice” for the use of this technology—not only with money to purchase tokens but also with their valuable data.
“Essentially, you are paying twice: once with money, and once more with something far more valuable, which is exclusive corporate knowledge, so that the AI intelligence can be useful,” Nadella revealed in a blog post, as quoted by TechCrunch. He concluded that the cost companies must “pay” will increase alongside the growing demand for AI performance: “The higher the performance you expect from an AI model, the more internal knowledge you must surrender to it.”