Microsoft (MSFT.O) CEO Satya Nadella said companies should treat powerful AI
models as potential internal threats and assume models may be compromised,
urging built-in "emergency brake" mechanisms so authorized personnel can pause
or shut down autonomous agents during task execution. He advised against relying
on a single model for critical decisions, called for immutable logs of agent
behavior, independent audits of systems, and mandatory disclosure of major
failures or security vulnerabilities including causes and details to help others
strengthen defenses.
