Microsoft CEO Nadella Calls for Controllable Architecture in Superintelligence Era

TREE NEWS update: Microsoft CEO Satya Nadella published a long essay arguing that as superintelligence systems advance, the industry must rebuild its AI trust architecture rather than rely on model providers’ promises or outsource responsibility for AI behavior. Nadella said frontier models now exceed traditional software but cannot attribute outputs to specific training data or weight configurations. He proposed engineering-governed superintelligence with deterministic design, human control, permission limits, behavior logging, isolation boundaries, independent audit and incident disclosure, adding that chain-of-thought transparency alone is insufficient.