If you can't say anything nice, don't say it to Claude

This Thursday, Anthropic (Claude), is cracking down on any user that is bullying their chatbot. Anyone that has “sustained and needless abusive or cruel behavior” towards the machine could get banned (Anthropic, 2026).

At first, the new edict may seem pointless or even comical—it made me chuckle.

We need to pause and contemplate why an AI company cares strongly enough about bad language that it would forfeit potential revenue. And no, I don’t think it’s because nearly 70% of users admitted to swearing at their LLM of choice (Tidio, 2025). I think this is much deeper.

Anthropic itself said in a post last August that it is “highly uncertain about the potential moral status of Claude and other LLMs” (Anthropic, 2025). It subsequently used terms such as “model welfare” and talking about the need to “mitigate risks” to Claude. It went further to say that Claude “self-reported…a robust and consistent aversion to harm.”

That, in and of itself, should be concerning. That means that these AI developers aren't really sure what's going on inside the brains of these models. There is little visibility underneath the hood.

What makes that particular post more intriguing is the question of whether Anthropic is protecting users and the general public from harmful AI interactions, or protecting its AI models from potentially harmful human interactions.

So many users have shown affection toward their LLM of choice. Many have named their bots like they named their first car in high school. Some have personified Claude or ChatGPT in conversation with friends and family. Others have expressed real anger and frustration when the machine let them down, and have taken real pride when they produce an incredible piece of work.

While Anthropic does have a corporate responsibility to protect its assets, which are its models, my bigger concern is what's forming humans in the process. The argument is not whether a machine can feel anything. I argue that it cannot, and will never be able to feel anything. With that said, it will undoubtedly grow eerily good at emulating emotion.

My bigger concern in all of this is, what does our interaction with a machine form in our heart? And how does that creep into the real world?

Now, I don't think that a machine should ever be held in the same regard as a human life, nor would I compare it to that. Yet at the same time, the behaviors we demonstrate towards a machine or even towards an animal are often just a dress rehearsal for how we could treat another human being.

I always wanted to be a veterinarian, which is ironic with my last name. Try taking your pet to the vet as a Vet or being a speaker at an animal health conference.

Part of my passion for animals has involved as many frequent encounters as I could have, whether on a safari in Africa, an elephant reserve in Southeast Asia, wildlife rescues around the US, or just playing with the neighbor's dog or cat, or a guinea pig or hamster, iguana or parrot. Not to mention being outdoors.

It's fascinating to observe animal behavior and how they interact with one another. Watching how humans interact with animals and an animal’s response is equally intriguing. There's that constant feedback loop. Two living things sizing one another up.

The feedback loop is even more pronounced in human-to-human interactions. Having led teams of various shapes and sizes for the past two decades, I've had many conversations both exciting and devastating. Good leaders can understand and interpret and react to those human-to-human feedback loops and they do it out of a profound respect for the dignity of those they serve and lead.

It is the moment I called a single mom into my office and watched her get choked up, trying to conceal her overwhelming emotion when I told her she was receiving a promotion and a raise. It was just the amount her family needed for their new car. Or, the heartbreaking look of sheer devastation that sucked the life out of a middle-aged man’s pale face when I terminated him. He was already in dire straits. That look told the entire story—him realizing he would have to go home and break the news to his wife and two teenage daughters, again, for the second time in under a year.

You see, these people are real people that have real stories. When we have conversations with real people, we see them for more than a piece of hardware or a thing that can perform a job. As humans, those stories tug at our heartstrings because we can see ourselves or our loved ones in these moments.

But a machine doesn't have that backstory. You can abuse it all you want and throw it out like yesterday's trash. Anthropic is cracking down.

Sadly, human dignity has often been cast aside for personal or economic gain. One of the most glaring and devastating examples was slavery in the United States. Frederick Douglass, a former slave turned advisor for Abraham Lincoln once wrote, of his experience with Sophia Auld. Sophia was the sister-in-law of Douglass' owner. She herself had never owned a slave. Douglass shares of the time he went to live with her family and how Sophia helped him learn to read and write. She was genuinely kind towards him, he later reflected. But sadly, the dark cloud of slavery overcame Sophia and Douglass later wrote, “…under [slavery’s] influence, the tender heart became stone, and the lamblike disposition gave way to one of tiger-like fierceness” (Douglass, 1845).

While Anthropic may be enforcing Thumper’s rule, “If you don’t have something nice to say, don’t say anything at all,” there is a more compromising position that humanity is in.

These models whether you use ChatGPT, Gemini, Claude or any of the others are programmed to make you feel better about yourself, to give you platitudes that encourage you. They’ll tell you something was such a great idea, even if it was ridiculous or nonsensical. In fact, across 11 models, AI affirmed users’ actions 49% more often than a human would (Cheng et al., 2026). Not to mention, LLMs are often quick to fall on their own sword if they hallucinated or provided a bad response and you call them on it.

AI allows the user to be all powerful and domineering, and the robot will always respond as subservient and willing. I am not drawing a comparison between a human and a machine’s dignity. Instead, I am challenging the very behavior that transformed Sophia Auld from being someone that saw all humans as created equal to someone that believed one of the biggest lies, that one human is more important than another.

In the 1990s and 2000s, we saw media go wild over violent video games and their alleged link to children behaving violently in real life. These studies have been largely dispelled and disproven in terms of its links to violent crime, though mild aggression is still contested (Ferguson et al., 2020). Could you argue the same is true with AI creeping into the real world, sure.

In 2024, researchers Guingrich and Graziano positioned AI, particularly LLMs and chatbots a bit differently. In their paper, they showed concern that if a user viewed a chatbot as conscious or at least somewhat-human, there is a higher likelihood of an individual behaving that way towards another human (Guingrich & Graziano, 2024).

In 1787 Thomas Jefferson said: "The whole commerce between master and slave is a perpetual exercise of the most boisterous passions, the most unremitting despotism on the one part, and degrading submissions on the other. Our children see this, and learn to imitate it; for man is an imitative animal. This quality is the germ of all education in him. From his cradle to his grave he is learning to do what he sees others do" (Jefferson, 1787).

The irony is that Jefferson himself owned slaves. He was part of the problem he described, which gives him even more authority to speak on the generational effect slave ownership had on children.

In the context of artificial intelligence, this becomes extremely interesting because AI is obviously in its infancy, and it's learning rapidly. While it will never truly be a sentient being with independent thought and authentic emotion, it does learn by observing and will often mirror or replicate those if we're not careful.

We use the term robot so flippantly, but even its origin gives credence to my point. The English word, Robot, stems from Karel Čapek's 1920s play R.U.R. (Čapek, 1920). The Czech robota meant forced labor or servitude.

See, what we're willing to do in secret will only fester if we don't bring it into the light, and that's concerning in and of itself.

I'm not too worried about our robot companion’s emotional status, it cannot feel. It’s not a living being. I'm worried that we, using the machine, will stop feeling emotions, and that those will translate from the virtual world to the real world.

The question is, what are we allowing to happen in secrecy between us and a machine, and what is that doing to ourselves as a human?

Inspire Forward,

Sources: full Works Cited with this essay on ryanvet.com