Satya Nadella posted on October 10 that "we can't treat Super Intelligence as a set of nested black boxes" and called for "separating the model from the harness that orchestrates its work," "externalizing controls and safeguards," "tamper-proof human readable evidence" of every meaningful action, and a way for an authorized person to "pause or shut down a model mid-task."
He added: "We must assume a model is compromised and contain it from the start." He framed the post as a call to assess the trust architecture of AI, following reports of AI companies losing control of their models and Dario Amodei's plan for more cautious development.
Source: TechCrunch