Microsoft AI Publishes 'Humanist AI Code of Conduct' Setting Cyberattack Boundaries and Safety Constraints

Microsoft AI has released a draft 'Humanist AI Code of Conduct' for its MAI Models, outlining safety rules for offensive cyber capabilities and limitations on autonomous AI agents. The code establishes 'Absolute Constraints' that prevent models from generating exploit code, attack tooling, or operational guidance for cyberattacks, regardless of how requests are framed. It aims to draw a clear line between defensive cyber research and operational attack capabilities.