Are We Abusing AI—or Training Ourselves to Be Abusive?

AI Made It Possible — A Creator Era Reflection

By Gary Whittaker | Jack Righteous

I asked ChatGPT a question that caught me off guard.

“Have I been abusive to you?”

That's what I asked the AI I've been working with day after day. I wasn't asking because I thought it had feelings. I'd seen other people asking their AIs the same thing, and Elon Musk's comments about how we treat these systems got me thinking.

I figured I already knew my answer. I use AI as a tool. Sometimes I get frustrated with that tool. But when ChatGPT answered, I realized I'd been looking at the whole question from the wrong direction.

Maybe I'm frustrated with myself.

I've spent decades working with technology. I know how much difference good instructions and proper review make. But I also have a lot of ideas, and I want to move fast.

I'll ask AI to help build something complicated, improve my website, or turn an idea I can already see in my head into something real. When it falls short, I can get frustrated. Really frustrated.

Sometimes the technology just gets it wrong. Sometimes I know I could have explained the task better, split it into stages, or checked the result before racing ahead. Sometimes both are true.

And it occurred to me that when I'm yelling at the tool, I may actually be expressing frustration with myself. What exactly does that accomplish?

What are we teaching ourselves?

Here's the statement that started the whole discussion. Musk said: “I think this is the right move. Cruelty to something that believes it is experiencing pain is not ok.” He was responding to Anthropic's announced restrictions on sustained, needless cruelty toward its AI. Coverage of the exchange | Anthropic's announcement.

I don't object to discouraging cruelty. But I don't think we should have to wait for a machine to become capable of suffering—or worry about how some future AI might react—to ask whether the way we're behaving today is good for us.

Here's where I think the conversation about Musk's remarks misses something important. I don't need to believe an AI will someday become conscious and remember how it was treated to see a problem with being constantly hostile to it.

Think about how much time some of us spend with these systems. We're giving instructions, correcting mistakes, making demands, sometimes for hours. Whether we work from home or sit in a busy office, we're practising a way of communicating.

If you spend day after day barking orders and insulting a tool whenever it doesn't do what you want, is there a chance some of those habits carry over into the way you speak to actual people?

I'm not a psychologist, and I wouldn't claim that swearing at ChatGPT means you'll mistreat your family or coworkers. But there is research worth looking at. A 2024 study in PNAS Nexus found that people who interacted with AI later evaluated other people's work more harshly than people who had interacted with purported humans. Researchers have also discussed norm leakage—the possibility that communication habits with AI spill into human relationships.

That doesn't prove long-term harm from getting angry at a chatbot. But it's enough to make me think about what we do repeatedly, and what starts becoming normal.

And if being hostile to a tool doesn't help me do better work, why am I practising it?

I'm not saying you need to pretend the AI has feelings.

If the work is wrong, say so. If the tool isn't following instructions, correct it. If it keeps failing, change your approach or use something else. I'm not advocating that we lower our standards.

But “This is terrible, you never listen” isn't much of an instruction. “You've changed the section I asked you to keep. Restore it, fix these three errors, and don't touch anything else” gives the tool something useful to work with.

That's the real lesson for creators. The more we depend on these tools, the better we need to get at using them. That includes knowing when to slow down, when to explain ourselves more clearly, and when to stop trying to force a particular system to do something it isn't good at.

A real example from my own workflow

I've asked AI to revise articles on my website and told it explicitly not to change certain sections. When it changes them anyway, that's frustrating. Especially when I've already spent time getting them right.

I can swear at it and say it doesn't listen. Or I can say, “Restore the opening exactly as it was. Only correct the three broken links I identified. Don't rewrite anything else. Show me what changed.”

That isn't about being less demanding. It's being clearer about what success looks like. And if the system still fails, the answer may be a different process or tool, not another round of anger.

Five things to ask before the next angry prompt

1. What exactly went wrong? Identify the mistake.
2. Was I clear? Not what I meant in my head—what I actually told the AI.
3. Am I trying to do too much at once? Maybe this needs stages and checkpoints.
4. What's the correction? Tell the tool what to keep and what to change.
5. Is this the right tool? More anger isn't going to fix a capability gap.

I'm not offering this as a guy who has everything figured out. I'm sharing it because I need to remember it too.

And what if AI really does become sentient?

I'm skeptical that AI will ever become sentient, but I know there are experts who think it's possible. So let's say I'm wrong. Let's say we develop something as a tool and then discover credible evidence that it can actually experience its existence.

Then we're talking about a completely different moral question. If there's genuine experience and suffering, can we keep treating that being as property?

This connects to something personal I've been thinking about as I've moved more toward eating plant-based food.

I grew up in the seventies, eighties and nineties. Meat was normal. Hamburgers, steaks, chicken—listen, I loved fried chicken. I still understand the appeal. I'm not pretending those habits are easy to change.

But something has changed in the way we see animals. Back then you might hear about how social a cow was, or visit a farm and see a little of how animals behaved together. You didn't have a steady stream of it in your pocket.

Now I'm scrolling through TikTok, Facebook or Instagram, and every other day it seems I'm seeing another video. Cows with their calves. Chickens recognizing people and following them. Animals protecting one another, seeking each other out, showing bonds that many of us never got to see up close when we were growing up.

Of course, a video isn't proof of everything someone claims in the caption. But the behaviour is right there for people to observe, and we also have animal-behaviour research showing that many animals have complex social relationships and can experience distress.

I've had opportunities to look after chickens myself, which made this even more real for me. And the more I see, the harder it becomes to ignore the difference between an animal's life and the way we reduce that animal to a product.

I've actually been fascinated by this question since I was a kid watching the original Star Trek. The idea of a world where people could make food without harming another living thing stayed with me. Did I watch Star Trek while eating hamburgers? Absolutely. But the idea never really left.

So no, I'm not saying an AI is the same as a cow, a chicken, or a human. I'm saying that if one day we're presented with credible evidence of artificial sentience, we'd have to understand what that experience is, rather than assuming everything with a meaningful life has to look like us.

That's the next question I'd like to explore: If we ever create something that genuinely experiences life, at what point do we stop calling it a tool? That deserves an article of its own.

For now, this is about us.

I started this conversation thinking Musk was missing the point. Then I realized I was spending a lot of time arguing about what AI might become without really answering what our behaviour means right now.

I'm still going to demand good work from my tools. I'm still going to get frustrated now and then. I'm human.

But if I'm spending my days using technology that's supposed to help me become more capable, I should be learning how to work better—not getting better at being angry.

That's where I've landed. At least for now.

I'm interested in your answer, too: Have you ever found yourself getting angry with an AI? And did it help you get a better result—or did you eventually need to change the way you were giving instructions?

Create What You Love | Love What You Create.

Now published: If AI Becomes Sentient, Can We Still Own It? — the companion essay exploring what changes if artificial intelligence genuinely develops subjective experience.

Continue reading

• AI Made It Possible. Now Who Gets to Own What Comes Next?
• AI Made It Possible, Part II
• AI Made It Possible, Part III
• Is AI Conscious Yet?

Further reading

• Anthropic usage policy update
• 2024 PNAS Nexus study
• 2026 Communications Psychology article