Microsoft CEO Nadella: Frontier AI Models Should Be Treated as Internal Risks with an "Emergency Brake"

PANews, October 11 - According to Bloomberg, Microsoft CEO Satya Nadella wrote that enterprises should treat powerful AI models as potential internal threats, assume they may have been compromised by default, and establish an "emergency brake" mechanism to prevent agent models from going out of control. He said enterprises deploying advanced AI should not rely solely on assurances from model developers; control measures such as permission management, operational restrictions, logging, and containment boundaries should be held by the deployer and placed in systems that the model cannot modify.