Microsoft CEO Satya Nadella called for new safety measures around artificial intelligence in a Saturday morning post on X, arguing that the industry needs to reconsider how much trust it places in the systems it builds. Nadella said it is time to step back and assess the trust architecture of AI. He framed the issue around what he described as Super Intelligence, the term the Trump administration favors for advanced AI.
Nadella outlined several specific changes. He said the model should be separated from the harness that orchestrates its work, and that controls and safeguards should be externalized. He also said every meaningful model action should be documented with tamper-proof human readable evidence. In addition, he called for systems in which an authorized person always has the ability to pause or shut down a model while it is in the middle of a task.
The Microsoft chief said a model should be assumed to be compromised and contained from the start. He compared the approach to an emergency brake. His argument is that advanced AI should not be treated as a set of nested black boxes whose recommendations, answers and actions are simply accepted or rejected.
The comments arrive as leading AI companies acknowledge a growing number of incidents in which they appeared to lose control of their models. Nadella is the latest technology executive to offer extended thoughts on how AI safety might be improved.
He is not the first industry leader to weigh in recently. Anthropic CEO Dario Amodei has published a plan for more cautious AI development. That plan and Nadella's post reflect mounting concern among AI developers about keeping advanced systems in check.
Nadella's proposal touches on accountability and oversight rather than model capability. The measures he described would place limits on what a model can do on its own and preserve a human's ability to intervene. For US technology firms and policymakers debating AI rules, his post adds a prominent executive voice to the argument that safeguards should be built into how models are deployed and operated.
More AI news from TechManNews.
