Microsoft CEO Satya Nadella has published a new post arguing that so-called Super Intelligence systems are black boxes that companies should not trust, and that strong deterministic systems are needed around them once they are deployed. The post, titled "Models as Insider Risks in the Super Intelligence Era," was shared from his account and picked up quickly by Techmeme and The Decoder.
The core claim is simple: today's models give no way to understand why they produce a given answer or decision, yet they are already wired into sensitive data and given the ability to take mission-critical actions. Nadella's answer is not a better model, but tighter control around whatever model is used.
Nadella's Super Intelligence warning, explained
"Not because they are necessarily malicious, but because any sufficiently capable actor with access to important systems can make mistakes or be compromised, and the architecture of containment and control must account for that," Nadella writes, framing models as insider security threats rather than external attackers.
His proposed fix is to separate intelligence from control over that intelligence. The most trustworthy system, he argues, is not the one with the best model but the one where you have to trust the model the least. He also points out that Google DeepMind has raised similar concerns about treating models as potential insider risks.
- Model diversity instead of dependence on a single provider
- Complete logging of every action a model takes
- Independent auditing of model behaviour
