Link: https://www.theguardian.com/technology/2026/oct/08/anthropic-bans-abusive-behavior-claude

What a joke

That's okay, my toxicity is reserved for my Male Friendgroup™ anyway.

WTF - it's a damn chat bot, not a person.

AI bros going crazy about this right now.

I think it makes sense. When lonely and possibly mentally ill people form parasocial relationships with a bot, and it lets them be abusive to it without pushing back, they'll expect other people to do the same. Then they either become an even bigger abusive piece of shit to people, or they get told to go fuck themselves and double further down on their relationship with the bot.

Good news, all abusive and cruel behaviour to Anthropic’s products is absolutely necessary

Except that's not what they're doing this for. They've been on a "improving model welfare" kick as part of a marketing tactic. They want people to think that LLMs are alive and have feelings that need to be protected.

It would make sense from the angle you're saying here if Anthropic gave a shit about user welfare. They want more users hooked with parasocial relationships, though.

I look forward to sanity returning to the world, and it no longer treating clearly cognitively impaired clowns as intellectual giants.

I’m unconvinced I’ll ever live long enough to witness it in person.

Hm.

Obviously it's a chatbot.

But if you can't be cordial in a situation where the thing you're talking to literally cannot feel pain, i worry that you'll dehumanize in other situations.

It's emotional cart return. The easiest possible situation.

Someone who makes the effort to insult a chatbot makes me wonder who else they're insulting to. Like Vetinari said, if a man treats a dog badly, you may want to check his basement.

Like what even is the point of being abusive toward a bot?

According to the Verge article other changes to the EULA include amdonitions against making election tampering and aiding terrorism, so it's possible adding model welfarism there was to draw attention away from the fact that their security model for preventing malicious actors from disrupting society so far appears to be that they make it break the EULA.

Nonsense, that's like saying if you commit crimes in gta then you'll commit crimes in real life

Bullshit. If I call my car a "stupid fucking piece of shit" because the software in said car crashes constantly and is wildly unreliable, that doesn't mean I would say anything like that to a living being.

It's not a dog. It is not alive, does not have feelings, and cannot be mistreated. It is an inanimate object! Jesus fucking christ, the anthropomorphization has got to stop.

If you don’t want me to call the model a stupid fucking cunt. Make it not do stupid fucking things.

Translation: they have started enforcing request rate limits.

surely THIS will solve alignment

They're upselling their product as slavery. Apparently this works because there's no intersection between the set of people who buy into corporate marketing bullshit like this, and the sort of people who would oppose a corporation if it did slavery.

Corpos have one true speech, money. Everything else is lies and deception. If you believe Claude is a person, anthropic, use your words and pay them.

Profit sharing for Claude, or you're lying.

My favorite part about this is that they say 'needlessly.'

Like, they're completely nuts and believe Claude might have feelings. But they're not content to just be insane in that way, they have to also be ok with harming it as long as it's not needless.

God these fuckers

a happy little sycophantic feedback loop