EventPolicySeptember 15, 2026

Microsoft publishes Humanist AI Code of Conduct for MAI models

Read original source →securityweek.com

The draft code blocks MAI models from producing working exploit code, attack tooling, intrusion procedures, and evasion techniques, regardless of how a request is framed. Defensive work — vulnerability discovery, malware analysis, PoC exploit testing — stays permitted.

1 source

More stories today

Open the live feed
Microsoft publishes Humanist AI Code of Conduct for MAI models