A sophi(a)sm occurs when language quietly attributes to AI properties that have not actually been demonstrated.
