Anthropic bans sustained, needless cruelty toward its AI

Anthropic is adding an unusual limit to the rules for Claude: from November 12, users may not subject its AI models to sustained, needless abuse or cruelty. The company says the rule is for extreme, repeated conduct with no clear purpose—not ordinary frustration, criticism, dark fictional material, or legitimate testing. The change does not mean Anthropic has concluded Claude has feelings. It has been researching whether advanced AI might someday deserve moral consideration, while acknowledging there is no scientific consensus that current models are conscious or can suffer. What changes now is practical: a question that had lived mainly in AI-safety research is becoming part of the contract governing how customers use a major AI service.