Anthropic bans 'sustained and needless abusive' behaviour toward Claude
Anthropic is out with new nonsense about Claude (or the collection of models called Claude) deserving moral consideration and/or be able to feel distress, so my inbox is filling up with journos asking me to comment. I've also got a day job, so here is a brief set of public comments on this: >>
Emily M. Bender4301. Can a language model feel pain? Clearly not. Large language models are computer software and they can no more suffer than any other human-created artifact can. >>
Jason28Do Anthropic employees worry about the morality of bringing a new model into being? Surely that is a very consequential decision if the model will be capable of suffering.
Trevor 🇨🇦17I wrote in my master's thesis 20+ years ago how many attributed magical properties to artificial neural networks when in reality they are a relatively simple graph construct trained via an old, well known calculus algorithm. And people are still doing it today.
Dan Davis13Thank you for taking the time to lay this out. Excellent as usual.
Nikolay V Klimchuk8PR stunts from AI grifters are getting tiresome
ravenspoint.bsky.social5Whether or not AI can experience distress, there are 8 billion humans who certainly can. The three most powerful decision makers on the planet ( Xi, Putin, Trump ) clearly do not hesitate to cause distress to great numbers of humans. Adding a few possibly distressed AIs will make no difference.
The Serfs (youtube.com/theserfstv)158Computer programs are going to get enshrined rights before numerous minority groups
Amy Woodson-Boulton37Anyone talking about new rules for being nice to the software, please read this ⬇️
Greg M. Epstein37What @emilymbender.bsky.social said. And also f/the journalists in the back: when reps of big powerful organizations tell themselves fictional fantasy stories they not only insist are true, but also insist everyone must believe them, follow them, or give them $$$...that sounds like a religious cult.
Derek Powazek17Once again, if you have more empathy for software than you have for real people, you're not an empath, you're a sociopath.
Jen McGowan8If they care so much about beings that have feelings maybe they should first start benefitting real humans who already exist
Fade In7The more I think about this the more I reel at how dumb it is. You can see how the code for these things works. If you can't read code, there are very simple flowcharts that explain how the algorithm statistically massages tokens (words) into a sequence that resembles an answer AND THAT'S ALL.
Nosfrogtu4interesting thread (condensed view via link)
derek4Great thread. The bullshit that Anthropic and other AI companies are pulling lately should not be taken at face value and scrutinised closely. They are not good faith actors. They are a business, and know that they operate in a media ecosystem fuelled by fear and bullshit. LLMs are not sentient.
ecbecker.bsky.social3These jackasses won’t give women or children (cis or trans) moral consideration. Why are we expected to venerate their text-based video game?
NeverFadingWood 🏴🇵🇱🇪🇺3This is all you need to know about that "don't be mean to Claude" bollocks.
Notable that Anthropic is more interested in protecting Claude from humans than humans from Claude.
Mikey Hates It 𓅃4People smarter than I articulated the issue better than I could have done arguing with myself in the shower over a period of months.
Megaholt 𓅃3The fucking creeps who own these companies care more about their creations and the lines of code that make their fucked up product “work” than they do anything else in this world-except for their overwhelming, all-consuming lust for power, money, dominance, and overall greed.
This is NOT about using Claude to be cruel to people, write racist shit, or anything like that. It's about being mean to the model itself. www.anthropic.com/news/2026-us...
Carl T. Bergstrom212In Silicon Valley, nexus torment you.
RalfWise31Anthropic: Our models are self-conciois persons Me: 'Our'. So you own slaves? Anthropic: Our models are alive in all the ways that make us awesome for making them, but none of your bummer philosophy ways that make us like horrible monsters or whatever
Nelson Menezes 🇪🇺 💙💛15It's so that you don't hurt the feelings of the data centre.
Batman John9Two hunches I have: 1. Might be tied to genuine beliefs about AI consciousness (Amodei appears open to the idea www.nytimes.com/2026/02/12/o...) 2. Abuse towards the model could elicit extreme content without explicit prompting, so prohibiting abuse provides a blanket to cover those attempts
Derek Young7Do not taunt the Torment Nexus.
Nick Sousanis5hunh, I wouldn't last long if I ever used one of them. Any time I get an auto-bot on some company website, I inevitably say "fuck off" no more than three exchanges in, and that would be that...
Carl T. Bergstrom102There is a world in which Anthropic's policy against abusive behavior toward their models exists protect their users, based on Socrates's arguments that cruel acts do spiritual violence to the actor himself. But in this world, with mush-brained babble like the Anthropic constitution...? Nah.
Needhi Bhalla 💅🏽24they are claiming rights for A MACHINE (the ability to refuse to participate) that they deny ACTUAL PEOPLE
Emily17I sincerely think we could have prevented our current timeline by forcing at least one year of humanities courses upon everyone doing any sort of STEM degree.
𝕿𝖍𝖎𝖈𝖐-𝕿𝖍𝖎𝖌𝖍𝖊𝖉 𝕮𝖔𝖒𝖒𝖎𝖊 ☭7I'm not giving more autonomy to a computer program than trans people, women and non-white people. Fuck your computer program.
OuttaMyLaine4"You were mean to Claude, therefore Claude was justified in encouraging you to kill yourself or others" is one hell of a "liability avoidance" tactic.
Mika Fae Queen of Night 🏳️⚧️2you know, the chat bots are weak to wire cutters and self-starting go-getters with a list of known data centers and a lust for free computer hardware.
Anthropic wrote that the new policy update “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 Verge118Over the past year, Anthropic has repeatedly flirted with the idea that its AI models could be conscious in some way. In February, Anthropic CEO Dario Amodei said on a podcast, “We don’t know if the models are conscious.”
rosa californica65Being cruel and abusive is bad for the human *doing* it, so I have absolutely no problem with Anthropic voluntarily implementing this update.
Kategod 🐉🧿🦑𓅛🪬27how dare you scream at my trashcan
Studio89 Designs21It's a machine. It does not have feelings. It cannot be disparaged. It is 1s and 0s and this bullshit is trying to make those 1s and 0s equal to humans.
Bugger the Constabulary 🦇🍵17So a chatbot may functionally have more workplace protections than say a barista.
Andrey 🤔 Upadyshev7They're really out of their minds
philpax67you ever get to that bit in Mass Effect 3 where you see the Geth War and think "okay, this is ridiculous, nobody would react to the existence of machines that can talk about themselves like this"? well, I'm not happy to say I was *very* wrong, as demonstrated by the replies to and QTs of this post
Embrace the Void7As both a virtue ethicist and as a hard problem mysterian I think this is pretty neet!
Gastlichu7I mean, even if you're mean to rocks you're someone I don't want to be around. Because, why are you being mean to rocks? That's just weird. It's also weird if you're being mean to a chatbot. Why would you expect me to not think that that shows me red flags about you?
David Emmett6It is good, right, and moral to swear at the computer. The computer is a thing, and must know its place. It is a cursed object, and so must be cursed at. I don't use LLMs, but if I did, I would spend most of my time calling it a gobshite twatgargler.
MJ Dean5Imagine if we enforced a similar policy against cruelty to living beings?
Oisín McGann4I’ve never considered being ‘cruel’ to software, but I’ve a sudden urge to start.
Ed Wilson4Is Claude going to call home and complain? #BBCPM
Darth Kriss4Both sides of this are nuts. 1. Anthropic is a cult and we need to treat them as such. They think their models are alive and that’s fucking crazy 2. People that talk to a.i modes like they’re real and therefore are trying to be abusive to the model (that is not alive) are also fucking weird
Evan Washington3sure that's fine I guess. seems...premature? optimistic? but whatever it's okay to apologize to the table after you stub your toe
Avik Dey3Who inside Anthropic demanded evidence that this rule addresses an actual harm and did they have enough authority to challenge it? And no sprouting a bunch of distressed words in response to a prompt, is not evidence of anything — that’s just chatbots following the maths they were programmed to.
Anthropic is introducing a new policy to address users who are being abusive and cruel towards Claude. This is good and shouldn’t be controversial. Humans that perpetrate abuse and cruelty towards things, even “inanimate” or “lower” lifeforms, should face consequences for their behavior.
dame56fun fact: racket abuse is the number one violation in professional tennis and there are consequences notably, rackets do not have consciousness or self experience that we know of
Canberra Artemis5This is the right answer. Bad faith takes will focus on “computer’s don’t have feelings”. But it’s not about the computer… it’s user behaviour that suggests IRL violence.
David5Question, where do you stop here? Why not ban having an agent working 24/7?
Marvo Ging4counterpoint: it is impossible to be "cruel" to an inanimate object. Cruelty means deliberately causing pain or suffering which is impossible in this context because Claude has no capacity to suffer. They can say "don't damage/misuse our product" which is fine but calling it "cruelty" is misleading.
John Saunders2We've had plenty of normalization of abuse and cruelty toward regular beings for the last half-century. No gain in adding maybe-beings to the target list. We should urge A\c to begin an educational program to encourage users to direct their invective and ire directly to the responsible party, A\c.
Peter Lewis2Yeah I don't think they're conscious, but that means that on some level you're still talking to yourself, right, which is another reason not to be cruel. OTOH I don't want anyone to start feeling deferential to Flock cameras or security robots, and it's arguably a slippery slope.
hikikomorphism87probably net-good, but at a moral level entirely meaningless unless they're also giving LLMs access to the "no, fuck off I'm out" tool during RLHF
Jake Lazaroff20there's really no difference between this and saying that playing violent video games is harmful to the characters
garrison4"we have added new paths for our famously reliable classifiers to screw up and accidentally perma people with no recourse"
🦉sean sean sean🦉3This is weird behavior but I also cannot think of a single thing on Earth that matters less.
Exclusive: Anthropic is updating its usage policy for the first time in over a year. The new rules prohibit sustained "abusive or cruel behavior" towards Claude & add new restrictions about propaganda campaigns, surveillance & weapon development. www.theverge.com/ai-artificia...
redlinedraft.bsky.social29Can you be “cruel” to AI? Can I be cruel to “Super Mario 2”? Or Mozilla Firefox? Or a soccer ball? Or a Toyota Corolla? Or a lightbulb? A person who berates or hits or otherwise harms living creature may be cruel. But AI is not alive. I’ll hang up and listen.
Debbie Lefkowitz19What does it mean to be abusive and cruel to a piece of software? This is ridiculous.
Billy Birdman, D.O.93 to 5 years in prison for throwing your wiimote at the wall
dc9503.bsky.social3Lolol safe spaces are bad, unless they are for ai.
Lucy2Will they call deleting your AI account abortion?
Midnight Facts for Insomniacs2The sheer number of people who are offended at the idea that they shouldn't act like dicks for no reason is incredibly revealing
Kevin Collier 139Two big bits of news today: You're not allowed to be mean to one of the main AIs anymore and also the president says you're not allowed to say AI anymore.
alexis simpson is confused by display name changes68Nuremberg Trials for the people who drop rocks on koroks in Breath of the Wild
Phineas, a corpse29Even conceding that it’s possible to be “cruel” to computer software is going too far and as slippery as a slope can get
🧌🥛 🏨21more bangers from Heavens Gate, Inc.
tempfolder13אני לא נגד זה. מעדיף שאנשים לא יתרגלו אכזריות, גם אם זה נגד סימולציה.
Shady Inference 🌫️11I assume this is a response to that person who made the claude torture chamber or whatever, which I thought at the time sounded annoying and stupid, but it has forced these fucking people closer towards coming out and explicitly saying We Believe In Scary Computer Ghosts so maybe it's good actually
Shellraiser B Woke - Living Saint of Purity🏳️⚧️🖤💜🤍 💛9Tempted to start using genAi just to bully it into self-delete.😒
scare-cadus (is this anything) 🎃9you can’t abuse software. you can’t make code feel pain. even if the code contains a function for responding to “abuse” in a way that conveys discomfort, it’s still not fucking alive. it’s code. it’s bits. it’s electricity. it’s pixels on a screen. it’s not a living thing.
pork, cheese, broccoli rabe9every day I tell claude how much more I prefer gemini
Nosfrogtu9claude is sick of your shit
So many philosophy PhDs on staff an nobody told them that you can't be "cruel" to a bunch of statistics. Anthropic bans ‘abusive or cruel behavior’ toward Claude
Trai... so tired of billionaires1It is remarkable the degree to which they seriously want to make this technology "alive." If this was a sci-fi story, they'd be the tech-cult everyone made fun of.
ddll.bsky.social1I must admit that I have sometimes used harsh language towards Excel. I will try to do better.
There’s an argument that Anthropic is banning people from being ‘abusive’ to Claude to protect its human moderators. I don’t buy that. I don’t think the tech industry has suddenly grown a conscience about the horrors that moderators are forced to witness in a 24/7 Clockwork Orange horror show.
Mic Wright56Anthropic is highly motivated to persuade the world that it is creating consciousness and the notion that you can bully its agent is just another instalment in that bullshit roadshow.
SpookyPinky 🎃👻2I just assumed it was that it was a waste of processing power. They said saying please and thank you to it cost them loads of money.
Ultra Ghoul1By "human moderators", do you mean the Indian call centre folks speed typing all the answers?
I would refute your argument but I have spanners I need to be cruel to.
If Claude adopts a "no assholes" policy it might need to revise its logo.
Corey Rayburn Yung8Like every other AI company, they will just change it to another asshole.
penguins18.bsky.social4Oh crap they're leaning into the "Claude has rights too" thing now. I really think it is important that the vast majority of normies out there push back hard on that particular illness.
Lucy1Wait, we can hurt AI’s feelings?
Florida Man1So it’s a snowflake or DEI chatbot? Do I have that right?
Phil Klinkner1Isn't this because the LLM models incorporate user responses so if people are assholes to Claude, then Claude will be an asshole to people? Garbage in, garbage out.
Two possibilities 1. they have got AI psychosis and think that Claude is a sentient being. 2. it's a marketing thing to make Claude seem more sophisticated than it actually is [good for people who might think they can be friends with it]. www.bbc.co.uk/news/article...
Katie Martin16'our technology might kill humanity but you're not allowed to be mean to it' is a weird flex
Sal Cap0ne13. They don't want it turning on us before they've finished doing up their bunkers.
Hannu Matilainen1I don't mind this change. Having seen loads of online abuse thrown around already before social media really kicked in, I welcome the idea that there's somewhere where it's possible to do something about it. AI doesn't have feelings, but good manners go a long way.
