Mustafa Suleyman warns against treating models as conscious

Mustafa Suleyman has published a warning against attributing consciousness, feelings, or rights to AI models, arguing that the evidence does not support such attribution and that doing so would complicate existing efforts at AI containment and alignment.

The statement, quoted by Simon Willison on September 16, 2026, frames consciousness as foundational to human ethical, legal, and political systems. Suleyman’s position, as cited, holds that extending any portion of these rights or protections to AI models is unjustified given current evidence, and that doing so would make containment and alignment work harder rather than easier.

Willison’s post, titled “Quoting Mustafa Suleyman,” presents the passage as a direct quotation drawn from a piece titled “A Warning About Model Welfare,” linked via citation to mustafa-suleyman.ai. The excerpt does not elaborate on what evidence Suleyman considers relevant, nor does it detail the containment and alignment challenges he references beyond the general claim that model welfare debates would compound them.

The quotation arrives amid wider public discussion of whether increasingly capable AI systems warrant moral or legal consideration. Suleyman’s remarks, as presented in this excerpt, stake out a clear position against that consideration, treating the question of model consciousness as unsettled by evidence and treating premature attribution of rights as a practical liability for AI safety work rather than a moral advance.