Microsoft Chairman and CEO Satya Nadella has urged enterprises to rethink AI governance, warning that advanced AI models present potential insider risks when granted access to critical corporate data and operational tasks.
In a post on X, Nadella argued that businesses must separate the generation of intelligence from the authority to act on it.
https://t.co/MjwFy73LOF
— Satya Nadella (@satyanadella) October 10, 2026
Because non-deterministic AI outputs cannot be easily traced back to specific inputs or training parameters, relying solely on developer assurances or internal self-verification creates “nested black boxes”.
To mitigate risks like errors, misinterpretations, or cyber exploitation, Nadella stated: “The most trustworthy Super Intelligence system will not be the one with the model we trust most. It will be the one that enables us to trust the model the least”.
He proposed seven principles for governing superintelligence: ensuring model diversity, enforcing complete observability with tamper-proof logging, practising continuous verifiability, keeping permission controls independent, maintaining independent auditability, establishing emergency containment shutoff mechanisms, and disclosing AI incidents transparently.
By placing deterministic software safeguards and human oversight around non-deterministic AI agents, Nadella emphasised that organisations can securely harness AI capabilities while maintaining complete operational control.
