Microsoft publishes provisional conduct rules for training its AI models

reported by 2 outlets· 2 articles · confidence: low · first seen 2026-09-14 16:27 UTC

What this means for you

Nothing to act on. There is no enforcement mechanism, no effective date and no published detail beyond general principles, so there is no way to check compliance yet. Treat it as a statement of intent until a version with obligations attached appears.

Microsoft published a provisional code of conduct for its AI models on Monday 14 September, setting out principles the company says its systems should follow. Mustafa Suleyman, chief executive of Microsoft AI, posted it on social media, writing that AI "must be subordinate and always in service of people". The code is intended to apply to the training of new models, and TechCrunch reports it tells models not to hack systems or trick humans. Neither account says who enforces it, what happens if it is breached, or when it takes effect.

Key facts

  • ·Microsoft published a provisional AI code of conduct on Monday 14 September 2026. source
  • ·It was published on social media by Mustafa Suleyman, chief executive of Microsoft AI. source
  • ·The code applies to the training of new AI models. source
  • ·Suleyman wrote: "AI must be subordinate and always in service of people." source
  • ·The code lists principles including supporting humans rather than replacing them, plus specific safety constraints. source
  • ·TechCrunch reports the code tells models not to hack systems or trick humans. source

What the sources say

  • TechCrunch AI — Sets out the document's general principles and the behavioural constraints it places on models.
  • The Guardian AI — Adds who published it, where, and frames it around company control of AI systems.

Sources

The original reporting. Follow these — they did the work.

← the wire