Anthropic will prohibit "gratuitous and repeated cruelty" toward its AI, Claude, starting 12 November, according to the supplied material. The rule may appear odd, but it reflects a shift in how laboratories approach the treatment of models: model welfare is being framed as a safety issue rather than only as a philosophical question.
The material presents the change alongside an open question — whether it amounts to a precautionary principle or carries a risk of anthropomorphism — and cites Thibaut Giraud, a philosopher and YouTuber, as the person addressing it. No further details of his position are provided.