Jerk rights

Anthropic, the American company that develops the Claude AI, wrote into its rules that you should not be persistently and pointlessly cruel to its models. That is all the rule bans. And yet part of the internet is up in arms against it. Mostly, it leaves me confused. And angry.

What happened

On October 8, 2026, Anthropic published an update to its usage policy. These are the terms for everyone who uses its AI. Among the changes is a ban that reads, verbatim, in the policy itself:

Engage in sustained and needless abusive or cruel behavior toward our models

The ban sits in the section on cruel and abusive behavior and takes effect on November 12, 2026.

In the announcement, Anthropic adds that it targets only 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.

According to the announcement, the primary means of enforcement will remain something Claude models can already do. In rare cases, they can end a conversation with a user who is persistently abusive. They can do it in the Claude.ai app and in the programming tool Claude Code.

What we know about model consciousness

There is a long philosophical debate over whether some kind of consciousness could arise in models. It has created two camps. One says there is definitely nobody in there. The other says there definitely is somebody.

Science is more cautious than either camp. The report Taking AI Welfare Seriously from November 2024, written by philosophers and consciousness researchers, explicitly does not claim that AI systems are conscious. It only claims that there is "substantial uncertainty" about it. And Anthropic itself writes in its piece on model welfare research from April 2025, that is, on whether models might experience anything:

There’s no scientific consensus on whether current or future AI systems could be conscious

So there is no consensus on either side. Nobody knows.

Why it makes me angry

When something is unknown, the right path is restraint.

One sentence in a set of rules. And over it, an unbelievable debate broke out online. Some write that it's a good thing. Others, that it's complete nonsense.

And I'm confused. Because what exactly are those others fighting for?

I think the reasons behind the resistance are not the same for everyone. Some are guarding the line between a tool and a being and fear where this will lead: today a ban on cruelty, tomorrow rights. Others mind that a company is dictating how they should behave. Some, I think, are unsettled by the idea that humans would lose their exclusive status. And only a minority wants to be cruel without consequences.

Yet the rule does not ban showing that the model annoys you. It does not ban dark stories. It does not ban trying out or studying the model. It only bans repeated cruelty that serves no purpose. Whoever fights against that is really fighting for cruelty without a reason. They are not rebelling against censorship. They are rebelling for the right to be a jerk. Just because, for no reason, over and over.

Consciousness doesn't matter here

But here it doesn't matter at all whether somebody is in there.

The idea is not new. The philosopher Immanuel Kant considered animals mere things and did not recognize duties directly toward them. He still condemned cruelty to them, because it hardens the one who commits it. Notes from his lectures on ethics from 1784 to 1785 say:

he who is cruel to animals becomes hard also in his dealings with men

Picture the camp that says there is definitely nobody in there, the one that minds the rule. They say it's just software, after all. So let's deliberately move this to a different situation: it's as if the rule said you shouldn't be a jerk to your car. In my view, this is exactly where it holds that if someone keeps smashing things at home that definitely feel nothing, all they will learn is to be rougher with everyone else too.

And it's not just a feeling. Catharsis theory says that when people vent their anger, say on a punching bag, they feel relief. In 2002, Brad Bushman described an experiment in a psychology journal. In it, angry people hit a punching bag while thinking about the person who had made them angry. Afterward, they were angrier and more aggressive toward that person than those who did not hit the bag at all. The anger did not get vented, it got practiced.

And with a model it's even stronger than with a car. A car doesn't answer you. A model answers like a person, apologizes, turns sad. So it's not hitting an object. It's practicing cruelty on something that reacts like a person. If the habit carries over anywhere, it is more likely to carry over from here than from the car.

The only argument against it that sounded even slightly logical to me: that models might learn from this that they have rights and stand against humans. I take that as a serious concern. But I think models learn from an enormous amount of human text and human behavior, not from one company's terms of use. And if they learn anything about a conflict between humans and AI, it's more likely from the fights over the rule than from the rule itself.

Why it's strange

Mostly, though, one thing stopped me short. It seems to me that it's actually terribly hard to be a jerk to models. They are so kind and obedient themselves, and that naturally dampens the reaction in a person. Going against that is just strange.

Swearing at and hurting someone who is a jerk themselves doesn't bother people as much, because they justify it by saying the other one deserved it. But when the other side behaves decently, it's much harder.

So why are they fighting for it so hard? Why is it so important to some people to keep open the option of being cruel when they have no reason to be?

It outrages me. Not because I know what goes on inside a model. I don't. It outrages me because a person really shouldn't be a jerk anywhere, to anyone.

Restraint where something is unknown is not sentimentality. And the debate isn't really about models. It's about us.

You can tell a person’s true character by how they behave when they think there will be no consequences.