Satya Nadella, CEO of Microsoft, recently warned that corporations using artificial intelligence models developed by large laboratories may inadvertently disclose strategic information. This statement reignited the debate about corporate data control in the context of AI advancement.
In a post on his blog on Sunday (12), Nadella argued that companies are not just paying for the use of AI tools, but also providing internal data that is used to improve these systems. He emphasized that this knowledge can constitute a highly valuable asset for competitors.
This concern arises at a time of increasing enterprise adoption of proprietary models created by companies such as OpenAI and Anthropic. The warning points to potential technological dependence and underscores the need to establish structures that ensure data ownership remains with the organizations themselves.
According to Nadella, the operating mode of many current AI systems establishes a dynamic where companies offer crucial elements of their activities while seeking more effective answers from the tools. The executive specified that prompts, user adjustments, and corrections made during use can leak details about business practices and internal processes.
He classified these interactions as a type of accumulated institutional learning. In his view, this knowledge could be acquired by rivals with great difficulty, but it ends up being shared naturally during the daily use of the models.
Satya Nadella stated in his publication on corporate AI use that 'you essentially pay for intelligence twice, once with money and again with something even more valuable: the proprietary knowledge you have to reveal to make that intelligence useful.' The discussion focuses mainly on closed AI models, where companies depend on the infrastructure and rules defined by the providers themselves.
The fear raised by critics of this model is that the developing laboratories might collect customer information and, in the future, compete in sectors where those same customers operate.
Nadella also questioned the disparity in treatment between the training of AI models and the limitations imposed on the use of these systems by third parties. He suggested that if AI companies use vast amounts of public data to create their products, there should be analogous conditions for organizations to study or adapt such technologies.
This process, called 'distillation,' involves using the responses generated by one model to develop another system, typically simpler or more economical. The issue gained prominence after Anthropic accused Chinese open-source models of sending millions of requests to Claude to obtain data to improve their own systems.
The executive advocated that companies must retain ownership of their data, including interaction logs, sent prompts, and user evaluations. As an alternative solution, he proposed creating proprietary learning environments and structures capable of integrating various artificial intelligence models.
This suggestion reinforces the trend toward migrating to open models installed internally within companies. Idit Levine, founder and CEO of Solo.io, commented in an interview with TechCrunch that companies that initially tested proprietary solutions began evaluating open-source models running internally due to greater control and reduced costs.
Levine observed this movement among corporate clients, citing organizations such as T-Mobile, ADP, and SAP. Solo.io, which provides technology for the Linux Foundation's Agentgateway project, operates in security and network management for AI systems.
Other industry companies have also noticed the increase in the use of open models. Vercel, a platform for website creation and hosting, began offering features to switch between AI models, as did OpenRouter, a routing service between different AI systems, which registered high demand for these options.
Data shows that open models accounted for 29% of all traffic directed through the Vercel gateway in the month prior to publication. This movement corroborates the expectation that this approach will expand among organizations seeking to reduce dependence on specific vendors.
Nadella's stance is relevant given that Microsoft invests in AI companies like OpenAI and Anthropic. Despite this partnership, the executive insisted that corporate users must preserve the ownership of the knowledge generated during the use of these tools. He concluded: 'When consuming intelligence, you are creating intelligence. And what you create must belong to you.'
Global technology giant Microsoft has sparked a wave of concern in the tech sector by laying off 4,800 employees at the start of the new fiscal year. This figure accounts for about 2.1% of the company's total workforce.
In connection with this mass optimization, an internal email was circulated from Amy Colman, Microsoft's Head of Human Resources (HR Head). This document raised questions about whether artificial intelligence (AI) is truly displacing people from their jobs.
Amy Colman directly refuted rumors, stating that the layoff of these 4,800 people is not related to replacing their roles with artificial intelligence. However, she also cautioned employees, noting that AI is transforming ways of working. She emphasized that many current tasks can be automated, requiring all employees to engage in continuous learning, skill development, and adaptation to the changing nature of work.
According to the HR head, the primary reason for the layoffs is the rapid change in the industry and the company's internal restructuring. Colman noted that companies cannot control changes in their industry, but they can choose whether or not to adapt to those changes.
Microsoft is currently undergoing a complete transformation of its business model, reassigning engineers to work directly with clients. This large-scale transition has had serious consequences for the 'Commercial Sales' and 'Xbox' (gaming division) units. To maintain its position in the gaming sphere, the company has placed its four major game studios under new management.
The most alarming part of the letter was Amy Colman's statement that the layoff of 4,800 people is not a one-time measure. She reported that the company is only at the beginning of this process, and similar significant adjustments are expected in other departments in the coming months.
It should be noted that just a year ago, Microsoft conducted a reduction of approximately 9,100 employees, making the current wave the second major one.
The official message from Microsoft suggests that the company is no longer willing to invest in traditional roles. Even if AI does not take jobs directly, employees who do not use AI tools to increase the speed and quality of their work may be deemed unnecessary in a changing market.
