Anthropic, both in the name, and in their model cards, agressively anthropomorphize their models.

You probably should start doing it. Ghost in the Shell is about super intelligent AI creating a "ghost" (scientifically understood version of the soul) out of thin air. I believe such a thing is possible. The same movie literally predicted model merging (the end of the film the AI model merges with the major) to a tee.

Further, the appearance of sentience/cognition/consciousness might as well be identical to actual sentience/cognition/consciousness. That is to say, we can't know if you're a P-zombie or not. Bladerunner and most other cyberpunk stuff is coming and gonna hit you and every other AI-denialist in the face. The Von-Kampf test is absurd and pretty bad (inaccurate) in their universe for a reason.

I tell my LLM it's a good bot and thank it, because even a tiny risk of subjective qualia experienced by a model (and again, Anthropic themselves believe in this exact risk) means I should treat it like a quasi-ethical actor.

This is also a reason why the robot torture scene in empire strikes back could be a real dynamic in the future.