Microsoft AI Code of Conduct Sets Cyberattack Boundaries, Chain of Command, Safety Constraints
Source SecurityWeek
Description
Microsoft has published a draft 'Humanist AI Code of Conduct' for its MAI Models, establishing strict boundaries to prevent AI from enabling cyberattacks. The code prohibits the generation of exploit code, attack tools, and operational attack guidance, enforcing absolute constraints that cannot be overridden by users or operators. It allows AI assistance for authorized defensive cybersecurity tasks such as vulnerability discovery and malware analysis. The code also mandates transparency in AI reasoning and restricts AI agents from escalating privileges or acting beyond authorized scopes. Special review processes are planned for cybersecurity and other sensitive domains. The draft is open for public consultation before finalization.