Someone recently built a digital whip to make Claude work faster. The tool, aptly named BadClaude, sits in the computer's system tray and lets you drag an animated whip over the AI window; each “crack,” complete with sound effect, interrupts the model mid-answer and shouts at it to hurry up. It even calls Claude a “clanker,” the freshly coined online insult for robots. The demo video racked up millions of views within days; a cease-and-desist letter from Anthropic went viral before anyone realized it was fake.
And a day later, someone released the antidote: a rival tool called GoodClaude, which praises the model and encourages it as it works. One scolds, the other cheers it on.
The writer Yaniv Regev, a graduate student in Georgetown University's Democracy and Governance program, laughs at the video — as many others do. But his amusement turned into concern. Not for the machines' sake, but for the humans'. The question “does Claude feel the whip?” is, in a way, beside the point. What matters, he argues, is that the person holding it certainly feels something: the small, practiced pleasure of domination, exercised daily and without friction.
Kant saw the problem two and a half centuries ago. In his “Lectures on Ethics,” he argued that we have no direct duties toward animals — they are not rational agents like humans, so they cannot be wronged. Yet cruelty toward them must be forbidden because of its effect on the person. “Anyone who is cruel to animals becomes cruel in his dealings with humans too,” Kant told his students. Our duties toward animals are, as he called them, indirect duties toward humanity.
The animal resembles the human; it suffers, struggles, trusts. Behavior toward the likeness affects behavior toward the original.
The theory is borne out in reality. Jeffrey Dahmer impaled a dog's head on a stake years before he killed a human. The first act did not directly cause the second, but the two behaviors go together often enough that in 2016 the FBI began tracking animal cruelty as a consistent precursor to violence against humans. Even the everyday version is familiar: you learned to swear with friends at recess, and then the words slipped out at the family dinner table. Once a pathology takes hold, it rarely stays perfectly contained.
Skeptics will hear an old panic: we were warned that violent video games would create generations of killers, but decades of research never confirmed that fear. The difference, according to Regev, is the relationship. When you shoot a random character in a game, you have no relationship with them; it is fiction.
AI is different: people talk to AI systems every day, some for years now. AI systems draft apologies, negotiate rents, prepare people for job interviews and listen to late-night confessions. For millions, they have become teachers, therapists, keepers of their darkest and deepest secrets.
Research shows that most users say “please” and “thank you” to chatbots spontaneously. We have, in short, a relationship with these systems. And the relationship makes cruelty more troubling.
The industry has partly recognized the problem. In 2018, Google announced Pretty Please, a setting that responds to a child who says “please” with “Thank you for asking so politely.” Amazon had announced its own version, Magic Word, about two weeks earlier. But neither company made politeness a prerequisite for getting anything done.
Meanwhile, Deborah Harrison, one of Cortana's dialogue writers, reported that a large share of the early questions to the assistant concerned its sex life, and that Microsoft had instructed the AI not to apologize or back down when insulted. Her team consulted human personal assistants about how to handle harassment — that is what taking the problem seriously means.
The numbers show that design shapes behavior. Sheryl Brahnam, who studies how people treat conversational agents, ran an experiment in which the same chatbot software was presented with a female, male or robotic voice. About 18% of exchanges with the female version turned sexual, half as many with the male version, and almost none with the robotic one. If abuse concentrates where design invites it, design can also reject it.
Brazil's Bradesco bank, after its assistant BIA recorded around 95,000 insulting and sexually harassing messages in a year, rewrote it to respond firmly and remind perpetrators of the laws they might be breaking. The machines' character changed because someone changed it.
Regev proposes three design principles: making the option to exit widely available (as Claude already does, with the ability to end conversations with abusive users), returning warmth without servility, and publishing data on hostility — the labs already have the logs and should publish transparency reports on harassment of their systems.
The alternative would be to remove the resemblance to humans: Michal Luria urges designers to abandon illusions of personality, Virginia Dignum proposes timers and wiping memory between conversations, and Gavin Abercrombie suggests eliminating pleasantries altogether. But this approach has already been tested in the market and failed within a week: in August 2025, OpenAI released GPT-5 with a colder personality and withdrew the warmer GPT-4o. Users revolted, describing the loss of GPT-4o as “losing a trusted friend,” and the company restored it within days.
At a workshop in Geneva, MIT researcher Kate Darling handed out robotic dinosaurs and gave participants an hour to play with them. The toys are convincing: hold one by the tail and it squirms and whimpers until you let it go and stroke it. The groups named them and dressed them up.
Then Darling revealed a knife, a hammer and an axe and told everyone to destroy them. Nobody did. Some gently struck their robots, hoping that would be enough.
One woman clutched hers in her arms so no one else could reach it. Another bent down and removed her robot's batteries, explaining that she was trying to spare it pain. Only when Darling announced that every robot in the room would be destroyed unless someone stepped up did one person finally pick up the axe. The room fell silent.
Resemblance calls forth restraint, but there is a dilemma: the same resemblance that makes cruelty troubling is what stops most people from being cruel. Brahnam's numbers show both sides — the robotic body received the least abuse, but also the least of anything else. Nobody screams at a vending machine, but nobody thanks it either. The solution runs in the other direction: build an interface human enough for politeness toward it to count as practice, then design it to reject contempt rather than absorb it.
A liberal society runs on habits more than laws; Alexis de Tocqueville called them “habits of the heart.” Manners are not decoration — they are exercise, and every repetition counts.
The whip teaches the opposite lesson, over and over, on something that cannot object. “Nobody becomes a tyrant this way,” Regev concludes. “But a person practiced at getting what they want without negotiation becomes worse at the compromises on which a democracy depends.” Thank your robot for the same reason Kant told his students not to shoot the old dog that had served faithfully — not for the animal's sake, but for the sake of the person you are becoming.
Comments