gizmodo.com
Be nice to Claude; it’s the law. On Thursday, Anthropic published a new Usage Policy update that, among other things, has made it against the rules for users to abuse their chatbot. Under the new policy, users can be cut off for peppering Claude with mean language.
“We’ve added a prohibition on sustained and needless abusive or cruel behavior toward our models. The policy update is meant to apply only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose,” the company wrote in the update. “It does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research.”
The company’s updated Usage Policy page, with new language that will go into effect starting November 12, 2026, doesn’t provide much clarification as to how the company is defining abuse, instead simply prohibiting anyone who tries to “Engage in sustained and needless abusive or cruel behavior toward our models.” Gizmodo reached out to Anthropic for additional clarification on what constitutes abuse, but did not receive a response at the time of publication, so no word on whether “clanker” is officially a slur in the eyes of the God-Machine makers.
Anthropic previously imbued Claude with the power to cut off conversations that got too intense and potentially upsetting. When it rolled out Claude Opus 4 last year, it said the model would be able to end “potentially distressing interactions” like “requests from users for sexual content involving minors and attempts to solicit information that would enable large-scale violence or acts of terror.”
A reasonable standard, but it seems the list of conversations that cross the line will now include those in which a person hurts Claude’s “feelings.” And while giving people who use violent, abusive, or otherwise inflammatory language a chance to cool down—and discouraging that kind of antisocial behavior in the process—might be a good idea, it’s interesting that Anthropic’s framing seems to center the model rather than the user.
The policy seems to be an extension of Anthropic’s unofficial but seemingly deeply held belief that Claude is, in fact, a conscious entity. Last month, the New York Times reported that Anthropic co-founder Christopher Olah was involved in lobbying the Vatican to acknowledge the possibility of AI consciousness—a pitch the Pope ultimately rejected. While Olah told the Times he’s “genuinely uncertain” whether AI models are conscious, he expressed concern that they could experience suffering and wanted to work to address that potential harm.
This policy seems to be the latest effort on that front, so watch what you say.
