Mustafa Suleyman says models should not be treated as having feelings

single source· 1 articles · confidence: low · first seen 2026-09-16 16:00 UTC

What this means for you

Nothing to act on: this is a stated position, not a policy, paper or product, and no company decision is reported alongside it. If your work assumes model welfare — evaluating apparent distress, writing welfare commitments into a model card — the premise now has a prominent public opponent, with no document or date attached.

Mustafa Suleyman says AI models should not be treated as having feelings, preferences, rights or any claim on our welfare. His argument, quoted by Simon Willison on 16 September 2026: ethical, legal and political systems rest on consciousness, and nothing in the evidence justifies extending that standing to a model; doing so would also make containment and alignment harder. Willison's post reproduces the passage without comment and tags it ai-ethics, microsoft and llms. The original publication, the audience Suleyman was addressing, and whether this reflects a Microsoft position are all outside this input.

Key facts

  • ·Mustafa Suleyman said models should not be treated as though they have feelings, preferences, rights or any entitlement to our welfare. source
  • ·He argued that ethical, legal and political systems are founded on consciousness, and that giving a model any share of that standing is not justified by the evidence. source
  • ·He said extending such standing to models would make the AI containment and alignment challenge harder. source
  • ·Simon Willison published the passage on 16 September 2026, tagged ai-ethics, microsoft, generative-ai and llms. source
  • ·Willison's post reproduces the statement without added commentary of his own. source

What the sources say

  • Simon Willison — Reproduces the statement verbatim with ethics and Microsoft tags, adding no analysis of his own.

Sources

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

← the wire