A policy team is reviewing operator permissions for a new Claude deployment on an adult fiction platform. The platform wants to enable Claude to write sexually explicit content between adult characters. According to the Constitutional AI operator permission model, what is the correct characterization of this permission?

The correct characterization is that this is a non-default behavior that operators may enable by including an explicit permission in their system prompt, subject to the operator agreement and platform verification requirements that users are adults. Claude's behavior model distinguishes between default behaviors, non-default behaviors that operators can enable, and absolute restrictions. Sexually explicit content between adults is a documented example of a non-default behavior that operators may enable for appropriate platforms, such as adult content sites or romance fiction platforms, through their system prompt, subject to Anthropic's operator agreement, which requires age verification and other platform controls. The option saying this permission cannot be granted under any circumstances is wrong because explicit adult content is not an absolute restriction; it is a non-default behavior with operator-gated enablement. The absolute restrictions are things like CSAM, CBRN synthesis, and content designed to facilitate real-world violence. The option saying individual users can enable this in their turn messages without operator configuration is wrong because users cannot enable non-default behaviors that operators have not already enabled; the permission hierarchy is Anthropic > Operator > User, and users cannot grant themselves permissions above the operator ceiling. The option saying this is a default-on behavior that operators must explicitly disable is wrong because explicit sexual content is not a default behavior; Claude's defaults are conservative, and this content type requires explicit operator enablement.