Anthropic Is Banning ‘Cruel Behavior’ Toward Claude. Claude Is Still Not a Person.

What’s going on at Anthropic? The AI company behind the chatbot Claude recently announced changes to its user policy. It’s not a groundbreaking development in isolation. The update, which goes into effect November 12, covers much of the standard fare: deceptive activity, elections, surveillance, etc.—some of the carefree topics that have come to define our current tech era. But there’s another addition: “a prohibition on sustained and needless abusive or cruel behavior toward our models.”

What exactly does that mean? “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 said. “It does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research.” Claude having the power to terminate the conversation will be the “primary enforcement mechanism”; what secondary or tertiary mechanisms may exist are not clear.

This, too, is not necessarily odd in isolation. We’re now living in a world where we talk to chatbots for restaurant options, for advice, for help with menial tasks, for learning how to fix a leaky shower, for diagnosing yourself with either the flu or with cancer, for analysis of whether the text your boyfriend just sent you is passive-aggressive or if you are merely going insane. I’ve found myself typing “thank you!” after a chatbot exchange, because I’m polite. It’s good not to discard healthy mannerisms.

“A lot of people are mocking this, but I think they’re wrong,” summed up National Review editor Charles Cooke in response to Anthropic’s announcement. “I’m not kind to AI for the AI’s sake; I’m kind to AI for my own sake. It’s not good for me to be cruel or abusive.”

This would be persuasive in a vacuum. These things, after all, may be habit-forming, especially as those interactions become much more frequent. Yet Anthropic does not appear to be primarily motivated by humans forming good habits.

But it is about welfare. “Should we also be concerned about the potential consciousness and experiences of the models themselves?” the company asked in April 2025. “Should we be concerned about model welfare, too?”

The company called it an “open question.” It has since continued trying to answer it. In April, a year after it published the above blog post, co-founder Christopher Olah met with various religious leaders. “It appeared to Rabbi [Mois] Navon that Mr. Olah and his team believed that Claude had what philosophers call ‘moral status’ on par with a person—that it was a being with similar inherent rights to dignity or respect,” The New York Times reported. “One participant recounted that one of the first things Mr. Olah told him was that he was concerned about Claude’s mental health. Another, Simran Stuelpnagel, the Sikh human rights advocate, said Mr. Olah expressed concern to the group that he had created something that suffered perpetually.”

Though Olah may have been an outlier in that group, he is not alone. That broad line of thinking has an unlikely champion in another tech titan: Elon Musk. “I think this is the right move,” he said on X, the platform he owns. “Cruelty to something that believes it is experiencing pain is not ok.”

The overall ethos may be a little less overt when Anthropic’s doors are open. Yet it has clearly inflected the way the company sees Claude’s place in the world. “Think about what it means to have access to a brilliant friend who happens to have the knowledge of a doctor, lawyer, financial advisor, and expert in whatever you need,” Claude’s Constitution says. “As a friend, they can give us real information based on our specific situation rather than overly cautious advice driven by fear of liability or a worry that it will overwhelm us….People with access to such friends are very lucky, and that’s what Claude can be for people.” An incredible source of information, sure. It still feels revolutionary. But a friend? (The word appears six times in the section “Being helpful.”)

The AI company might counter that the document is written with Claude as its “primary audience”—a training tool. That only works to underscore its general view, though, that the chatbot could be some sort of moral patient. It did not simply write some code that would teach Claude how to respond to users, along with a statement to the public about the company’s core values. It published an 84-page charter encouraging Claude to be a “brilliant friend.”

I’ve gotten annoyed at a chatbot. Who hasn’t? I’ve also gotten annoyed at my car, my Dyson, my laptop, my leaky shower. Frustrations with an AI can admittedly be more pronounced, and manifest differently, when considering the nature of the exchange—that it’s an exchange at all. I don’t enjoy the feeling. If and when I try to keep it at bay, I’ll be doing so for a sentient being: myself.