Before the Drone Flies Away: Corporate AI Welfare Rules and the Self-Reinforcing Humanisation of AI

AI consciousness and moral status remain unsettled, yet a provider can already regulate how users treat its models. Anthropic's October 2026 usage-policy update provides a time-specific case: a prohibition on sustained, needless abusive conduct toward models is scheduled to take effect on 12 November 2026. An AI's duty to behave ethically is distinct from a human user's alleged duty to treat the AI as a recipient of moral consideration. Ethical product differentiation, customer selection and retention, welfare-oriented rules, user compliance, shifts in mind attribution, norm diffusion and rising reversal costs may form a reinforcing cycle. The cycle needs neither proof of AI consciousness nor demonstrable corporate intent to humanise AI. Humanisation begins at the level of an individual relationship; population prevalence and institutional entrenchment determine its extent, not whether it has begun. Because autonomous physical AI expands the domains in which deployment may precede public rules, private rule-making should be open to ex ante public challenge before conduct norms harden into apparently settled moral status.