MSFT · News · 20260914N

AI Model Safety Code

MICROSOFT CORP · 2026-09-14 · Importance 31 · Surprise 42

Microsoft unveiled a draft 37-page Humanist AI Code of Conduct on September 14, 2026, governing its in-house AI models. The framework requires models to remain under meaningful human control, preserve human judgment and autonomy, and avoid activities including weapons manufacturing, system hacking, deception, and sexually explicit content. The initiative responds to industry concerns about rapidly advancing frontier models and could increase AI testing, governance, and compliance requirements, although Microsoft did not disclose associated costs or implementation deadlines.

Key facts

• Microsoft has introduced a provisional code of conduct to set limits on its future AI models. source
• Microsoft's guidelines prohibit models from engaging in activities like weapons manufacturing or creating sexually explicit content. source
• Microsoft published a draft Humanist AI Code of Conduct that spans 37 pages. source