Policy move makes AI welfare language official

The company has enshrined the rule — summarized in Korean coverage as "no harsh treatment of AI" — in its official usage documentation, turning what had been a informal stance into explicit policy. The provision bans users from directing models to engage in abusive or harmful interactions.

Anthropic, the maker of the Claude family of large language models, has been among the AI firms most vocal about the possibility that advanced AI systems could merit moral consideration, though the scientific question of whether AI can suffer remains unresolved.

Significance for the industry

The codification marks a notable step in an emerging debate over "AI welfare" — how developers should treat increasingly capable models. By placing the restriction in its usage terms, Anthropic gives itself a contractual basis for restricting accounts that abuse its systems, alongside existing safety rules covering harmful content generation.

The company did not attach specific enforcement penalties to the provision in the material reviewed, and it remains unclear how actively the rule will be policed in practice.

Anthropic's stance stands out in an industry where most rivals frame model restrictions purely in terms of user safety and legal compliance rather than the model's own treatment. The claim that Anthropic has made a ban on AI mistreatment explicit in its policies is True.