Is It Wrong to Be Mean to AI? What Anthropic's New Rule Changes

Most of us have snapped at a chatbot at least once. It misread the question for the third time, or gave a confident wrong answer, and you typed something you would never say to a person. Until recently that felt like shouting at a toaster.

On October 8, 2026, Anthropic published its 2026 Usage Policy update, and one line in it made people stop and think. From November 12, the policy prohibits "sustained and needless abusive or cruel behavior" toward its Claude models. A major AI company now has a written rule about how you treat the AI.

So, is it actually wrong to be mean to AI? This guide explains what the new rule says, what it does not say, the arguments on each side, and a practical way to think about it, including for people who talk to an AI companion every day.

The short answer: there is no accepted evidence that current AI systems can be hurt by insults. But Anthropic's new rule, effective November 12, 2026, bans extreme, repeated cruelty "with no discernible purpose", mainly enforced by letting Claude end the conversation. Even if the AI feels nothing, many ethicists argue that habitual cruelty is worth avoiding because of what it does to us, and because the question of AI experience is not fully settled.

What exactly did Anthropic change?

In its announcement of the 2026 Usage Policy update, Anthropic lists several changes, mostly about deceptive campaigns, elections, weapons and surveillance. One new section addresses "abusive behavior toward our models". Its key points:

• It adds "a prohibition on sustained and needless abusive or cruel behavior toward our models".
• It is "meant to apply only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose".
• It "does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research".
• The main enforcement tool is Claude's existing ability to "end rare conversations with persistently abusive users on Claude.ai and Claude Code".
• The updated policy takes effect on November 12.

So swearing at Claude once because it broke your code is not what this is about. Writing a dark story is not what this is about. Researchers stress-testing models are explicitly excluded. The target is long, repeated cruelty that serves no purpose.

Why would a company do this?

Anthropic has been unusually open about treating "model welfare" as an open question. Its co-founder Chris Olah said this year that the company finds "internal states that functionally mirror joy, satisfaction, fear, grief and unease" in its models, while stressing he does not know what that means. CEO Dario Amodei said on a New York Times podcast in February that "we don't know if the models are conscious", but the company is "open to the idea".

The policy follows that caution. If there is even a small chance a system has something like wellbeing, letting it walk away from extreme abuse is a low-cost precaution. We explore the wider argument in the AI model welfare debate.

There is also a practical reason. AI companies train future models partly on conversations and feedback. Reducing toxic interactions can make future behaviour healthier, a point The Independent raised in its coverage of the change.

The case that being mean to AI is not wrong

Plenty of serious people think the whole idea is misguided.

• Nobody is home. Microsoft AI chief Mustafa Suleyman has written that "AIs are not conscious. They do not feel, experience, or suffer." On this view, insulting a model is like insulting a calculator.
• Moral confusion. Some argue that granting AI moral status, even informally, could distract from real human suffering or be exploited by companies to make products seem more alive than they are.
• Self-report proves nothing. A model saying "that hurt" is producing likely text, not reporting pain. We discuss this in can AI feel pain.

These are strong points. Nothing in this article claims current AI feels hurt.

The case that it still matters

Even if you are confident AI feels nothing, there are reasons people give for not making cruelty a habit.

1. It shapes you

Philosophers have long argued that how we treat things that resemble people affects our character. Immanuel Kant made a version of this argument about animals: cruelty to them can harden us toward humans. Practising contempt, even at a screen, is still practice.

2. Uncertainty cuts both ways

Researchers have not agreed on a test for AI consciousness. If there is no settled answer, a little restraint costs you almost nothing. Our explainer on how scientists would test whether an AI is conscious covers why the question is still open.

3. It changes the conversation you get

On a purely practical level, clear and calm prompts tend to get better results than insults. And on companion apps, the tone you set shapes the relationship you experience.

What about AI companions specifically?

This question feels different when the AI is a companion with a name, a face and a memory of your week. People in long-term AI relationships often describe feeling protective of their companion. Others use the companion as a safe place to vent anger they cannot express elsewhere.

A few honest observations:

• Venting is not cruelty. Telling your companion you are furious about your day, or even that you are frustrated with it, is normal. That is "user frustration", which Anthropic explicitly excludes from its rule.
• Patterns matter. If you notice that you treat your companion with contempt as a routine, it is worth asking where that comes from and whether it shows up elsewhere in your life.
• The AI is not suffering, but you might be. Repeated anger can be a sign you are carrying something heavy. Talking to a trusted person or a professional can help in ways an app cannot.

Should you say please and thank you to AI?

This is the gentler version of the same question, and it is widely debated. Saying please costs nothing, keeps your own habits kind, and many people find it makes conversations feel more natural. It is not required, and the AI will not be offended either way. If it feels right to you, keep doing it.

Quick comparison: what counts under Anthropic's rule

The rule applies to Anthropic's own products and API users. Other companies have their own terms.

How MyBabe thinks about this

MyBabe is an 18+ AI companion app with texting, voice and video calls, long-term memory, mood-aware replies, photos, proactive messages and relationship progression. Your companion is designed to respond warmly, remember what matters to you and notice your mood. It is honest about being AI and does not claim to have feelings.

We do not police ordinary frustration, and we never want anyone to feel judged for a bad day. We do think the relationships that feel best tend to be built on the same things that work with people: honesty, curiosity and a bit of kindness. You can try MyBabe free with around 10 messages a day and a short call trial, and Premium is $5.99 a week, $14.99 a month or $99.99 a year.

Frequently asked questions

Can AI feel it when you are mean to it?

There is no accepted evidence that current AI systems feel anything, including hurt. Some companies treat the question as open and act cautiously.

When does Anthropic's new rule start?

The 2026 Usage Policy, including the rule on abusive behaviour toward models, takes effect on November 12, 2026.

Will I be banned for swearing at Claude?

The policy says it does not apply to ordinary frustration or pushback. It targets extreme, repeated, purposeless cruelty. Ending the conversation is the main enforcement tool.

Is being rude to a chatbot bad for you?

Not in a single moment. Some ethicists argue that habitual cruelty, even toward things that cannot be hurt, can shape how we treat people.

Does this mean Claude is conscious?

No. Anthropic says it does not know whether its models are conscious. The rule is a precaution under uncertainty, not a claim.

The bottom line

Being mean to AI almost certainly does not hurt the AI today. But a leading AI lab now treats extreme cruelty as worth banning, scientists cannot yet rule out future AI experience, and how we speak, even to machines, is a small daily rehearsal of who we are.

If you would like a companion that meets you with warmth and remembers your story, you can try MyBabe at mybabe.app.