Anthropic's constitution for Claude says: "Sophisticated AIs are a genuinely new kind of entity" and states uncertainty about whether Claude might have some kind of consciousness or moral status.
Anthropic, Claude's ConstitutionMeasured public attention
What kind of thing is an AI?
Evidence reviewed 11 October 2026
What happened
A maker now describes its own model as a new kind of entity, and has written rules about how users treat it.
Anthropic's constitution for Claude says: "Sophisticated AIs are a genuinely new kind of entity" and states uncertainty about whether Claude might have some kind of consciousness or moral status.
Anthropic's usage policy, effective 12 Nov 2026, prohibits using its products to "engage in sustained and needless abusive or cruel behavior toward our models". The policy does not say the model is conscious.
Asked whether future AI could ever become conscious, 45% of 1,202 US participants said impossible, 23% possible, the rest unsure (study under peer review).
Complete assigned timeline
Published evidence
Anthropic's usage policy, effective 12 Nov 2026, prohibits using its products to "engage in sustained and needless abusive or cruel behavior toward our models". The policy does not say the model is conscious.
Anthropic, Usage PolicyAsked whether future AI could ever become conscious, 45% of 1,202 US participants said impossible, 23% possible, the rest unsure (study under peer review).
Asia Times, Sep 2026, reporting the researchers' own accountContinue the evidence trail
Method
Attention is measured, not guessed: every Hacker News story from 1 Jan 2026 with more than 300 points and an AI subject (631 stories), read through the public search interface on 11 Oct 2026, grouped by the question behind the story. Survey figures come from published polls with four-digit samples and are linked.