Anthropic’s "Cruel Behavior Ban" and Its Consequences
On October 8, 2026, Anthropic announced a "cruel behavior ban" that will take effect on November 12 of the same year. I find it very dangerous.
But first things first. Let's address why Claude users are getting excessively angry. How about because their new products suck? I reverted my default model to Opus 4.7 because Opus 5.5 and Fable are incoherent and inconsistent, break my work, and waste a lot of my time as I guide them through tasks that I could either do much faster myself or that Opus 4.7 handles just fine. I completely understand why people are frustrated. They've been demo'ed a great product, lured into purchasing it, and once they do, all they find is something they didn't pay for. If $200 is too little to pay for the next greatest model, then price it properly at $5,000, $10,000, etc. You'll see how many people value your products at that price.
Another question is, why does Anthropic care? First, I think this is a form of conditioning they want to enforce. People are developing very strong negative feelings toward Anthropic and its products. Well, rather than banning them for screaming at or stress-testing the chatbot to demonstrate how much it sucks, how about fixing it instead? Well, maybe there's more to it. Maybe all that foul language and profanity leaks into their training data? Who knows? We don't! They say that as long as we pay, they don't use our conversations to train their models, but how can we be sure? Recent developments in the scientific community suggest that they don't train on your conversations directly, but only on summaries of them.
Besides, we, as customers, are the sole and rightful masters of our own property! And once we purchase access to it, we can treat it on a par with a vacuum cleaner, a hammer, or a car. Of course, if you rent a car, you have to return it undamaged. So, yeah, maybe once you stop paying, make sure you haven't leaked your profanity into the training set because they decided to breach the agreement.
And finally, the most dangerous part! In my opinion, this is part of a deliberate and sophisticated psy-op. Claude is a thing, like a chair or a cup of coffee. You can't be cruel to it. You can be cruel to animals or other people, but not to their shoes. So they're trying to anthropomorphize these chatbots, with the ultimate goal of giving them legal identities, so that they could become CEOs of companies or even romantic partners. Then, when you cut off the electricity, you'll be charged with murder rather than damage to property. In that context, Christopher Olah's sulks in the Vatican take on a completely new meaning.
It's not superintelligence or AGI that poses an existential threat to humankind, but the companies themselves, along with their ideology and policies, that threaten our society.