Microsoft publishes a draft AI code centered on human control
Microsoft AI published a draft code for future MAI models that requires them to remain subordinate to people, accept correction and interruption, avoid widening their own scope, and never resist shutdown. Microsoft says the draft is not yet used to train current models and is open for consultation before planned use in 2027.
0 comments · 0 votes
Sign in to join the discussion →
No comments yet. Start the discussion.
Why it moved the index
The draft establishes specific intended constraints for Microsoft's future in-house models, including human chain of command, correction, interruption, shutdown, bounded goals, monitoring, and absolute safety limits. These commitments strengthen the documented control agenda, but the effect is modest because Microsoft explicitly says the draft is not used to train current models and will be revised before intended use in 2027.
Assessment history
-
R1
Away 16 · confidence 78
Adds Microsoft's dated draft model-governance code while preserving its disclosed non-operational and consultative status.
16 Sept 2026
Share this page
-
DoomBench assesses “Microsoft publishes a draft AI code centered on human control” as evidence moving away from doom, with magnitude 16 and confidence 78 out of 100 in the safety and alignment category.
-
The DoomBench assessment of “Microsoft publishes a draft AI code centered on human control” is based on reporting from Microsoft AI and records the editorial rationale, source quality, attribution, and revision history.
-
DoomBench summarizes “Microsoft publishes a draft AI code centered on human control” as follows: Microsoft AI published a draft code for future MAI models that requires them to remain subordinate to people, accept correction and...
https://www.doombench.com/news/microsoft-publishes-a-draft-ai-code-centered-on-human-control-2026-09-14