Microsoft CEO Satya Nadella has adopted the term Super Intelligence in his recent blog post, aligning with terminology preferred by President Trump. This shift replaces the traditional Artificial Intelligence label, despite the term having a distinct scientific meaning. Nadella uses this framing to highlight risks associated with model developers like OpenAI and Anthropic.
Nadella argues that AI models should be managed as potential insider security threats. He claims that current systems lack transparency regarding their decision-making processes, yet they are granted access to sensitive data and mission-critical operations. He suggests that companies must account for the possibility of AI systems making errors or being compromised.
To mitigate these risks, Nadella proposes separating intelligence from control. His recommendations include ensuring model diversity, implementing comprehensive action logging, conducting independent audits, and maintaining the ability for human intervention to shut down systems. He asserts that the most reliable architecture is one that minimizes the need to trust the model itself.
The strategy appears aimed at eroding trust in OpenAI and Anthropic, which are competitors to Microsoft's business interests. Microsoft has previously criticized these companies over data usage and model distillation. Meanwhile, Microsoft has been replacing third-party models in its Office suite with internal alternatives to protect its profit margins.
Microsoft faces challenges as AI models evolve into primary computer interfaces, threatening its position as the default interface provider. The company is currently reliant on external frontier models and struggles to pivot its legacy software business. Nadella's focus on security and control serves as a strategic effort to manage the rapid industry shift.
