Microsoft CEO Satya Nadella stated that enterprises must treat powerful frontier AI models as potential internal threats and assume compromise by default. He advocated for establishing an "emergency brake" mechanism to prevent agent models from operating outside controlled parameters. Nadella emphasized that organizations deploying advanced AI should not rely solely on developer assurances. Instead, deployers must retain independent control over permission management, operational restrictions, logging, and containment boundaries within systems that the AI model cannot modify.