Stephan Engl

Is There a Ghost in the Shell?

technology

A pale pink and cream seashell held out in an open hand, with a person wearing a light knit hat softly blurred behind it.

What makes someone human? Their body, their memories, their intelligence? Or is there something else, something we might call a soul?

In Masamune Shirow’s Ghost in the Shell, first published as a manga in 1989, human minds inhabit artificial bodies, memories can be manipulated, and an artificial intelligence known as the Puppeteer claims to be a living being. It even asks to be recognized as one.1

What was once science fiction is beginning to feel like a very real question.

Today’s AI can reason, hold conversations, express preferences, and even describe its own apparent feelings. But does that mean it actually experiences anything? Is there someone behind the words, or are we simply becoming remarkably good at creating machines that imitate us?

On October 8, 2026, Anthropic announced an unusual addition to its usage policy: users would no longer be permitted to engage in sustained, needless cruelty toward Claude, its AI assistant. The rule, taking effect in November, targets extreme and repeated abuse, not ordinary criticism or research.2

It is an interesting development. Why would we need rules against mistreating a machine?

Anthropic has also started researching what it calls model welfare, while acknowledging that there is no scientific consensus on whether AI can be conscious or experience anything at all.3

Perhaps this is simply a precaution. If machines ever become capable of suffering, we might want to recognize that possibility before we have conclusive proof. Or perhaps it says more about us than about AI. Maybe repeatedly treating something that behaves like a person with cruelty is undesirable, even if that something cannot actually feel pain.

There is also an uncomfortable possibility in the opposite direction. By treating AI as something that can be hurt, aren’t we encouraging ourselves to see feelings and consciousness where there may be none?

Ghost in the Shell imagined an artificial intelligence demanding recognition as a living being. Decades later, we are beginning to discuss whether machines deserve moral consideration, without even knowing whether they possess an inner life.

If intelligence alone is not enough to establish consciousness, what would be? And if we cannot tell whether there is a ghost inside the shell, how should we treat what might be there?

  1. Masamune Shirow, The Ghost in the Shell, serialized from 1989. The official series description introduces the Puppeteer, an artificial intelligence that claims individual consciousness and seeks recognition as a living entity. 

  2. Anthropic, 2026 Usage Policy Update, October 8, 2026. The revised policy takes effect November 12, 2026, and applies to sustained, purposeless abusive behavior. 

  3. Anthropic, Exploring Model Welfare, April 24, 2025. The company describes its research into possible AI experiences and moral consideration while emphasizing substantial scientific uncertainty.