Claude shows human-like 'global workspace' inside, Anthropic research finds

The piece frames the Claude finding less as a milestone than as a question about categories. Human brains constantly run vast processing that the person never notices — sudden ideas, unexplained likes, hunches — and only a fraction surfaces as "what I thought." If Claude shows a similar layering of internal processing, selection and output, the boundary between "humans have consciousness, AI just calculates" grows harder to draw.

The research itself stays cautious: what was found resembles access consciousness, not the subjective feeling of pain or joy. But the author's larger point is that human consciousness has the same proof problem — we infer other minds from speech and behavior, never directly. As AI systems accumulate memory, self-recognition, internal states and eventually first-person reports, the grounds for saying "but you have no real consciousness" become less obvious.

So the unsettling conclusion is not that AI is becoming human. It is that studying AI keeps returning the question "what is a human?" — and that the moment consciousness arrives may never be a clean, detectable event, but a gradual piling-up whose boundary we only notice in hindsight.