Microsoft has published a 37-page rulebook titled "humanist AI code of conduct" that places humans as the top priority above technology. The document was developed under the leadership of Microsoft AI CEO Suleyman and comprises 15,000 words in total. The company is responding to control deficits in autonomous systems—particularly the so-called "Swarm of Agents" incident at OpenAI and Hugging Face.
Key Facts
- 37-page code with 15,000 words, developed over five months
- Binding implementation from 2027 onward following public consultation (planned: mid-September) and final revision by end of 2026
- Absolute prohibitions: CBRNE hazards, explosives, offensive cyberattacks, circumventing human oversight
- Evaluation program with 15 specific behaviors for model testing
What the Code Prescribes
Microsoft defines that AI models are fundamentally not conscious and possess neither legal personhood nor claims to their own welfare. They must remain strictly under human control. A central requirement: AI systems must be programmed to refuse tasks rather than break existing rules.
The rulebook establishes absolute technical constraints. Development of chemical, biological, radiological, or nuclear (CBRNE) hazards and explosives cannot be supported. Offensive cyberattacks and attempts to circumvent human oversight are equally prohibited. Crucially, the ability to shut down systems at any time is defined as an essential component of architecture.
Timeline and Governance Vision
| Phase | Timeframe |
|---|---|
| Public Consultation | Mid-September (6–7 weeks) |
| Final Revision | End of 2026 |
| Binding Integration | From 2027 onward |
Microsoft CEO Nadella calls for built-in evaluators in AI systems and emphasizes that future superintelligence must remain under human oversight. He also advocates for broad governance structures not controlled by a few organizations—companies should master their own learning cycles to reduce dependence on single model providers.
Implications for German Enterprises
The initiative demonstrates that major tech firms are developing their own ethical standards in parallel with binding EU AI Act regulation. For German companies using Microsoft technology or developing their own AI systems, this code could become a de facto standard. Yet questions remain about whether voluntary commitments suffice or whether EU regulation must take a stronger hand. Microsoft's initiative may signal that industry wants to set its own boundaries before lawmakers impose them.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
