The Bourbon Defense

I was talking with a close friend this weekend as we looked out over the Strait of Juan de Fuca and watched seals play in the waves. He is going through a rough time. He's getting divorced, and it isn't a pleasant situation. Right now, my friend hates lawyers, and that's all he can think about. This hatred was reinforced when I commented that the seals seemed happy, and he responded, "Of course they're happy. They have the ocean, the waves, the kelp, and the clams, everything they need, and no lawyer in the waves following them around saying she gets half of everything, she gets half of your life, deal with it. Of course, they're happy. There are no fish lawyers. And in general, I don't believe fish are litigious." Then we silently watched the seals and the waves and drank our coffee.

He's not wrong. Fish are not litigious. We, on the other hand, are. We are land dwellers, and we sue. We talked about the recent story of a woman who had fifteen drinks on a cruise ship, fell down, injured herself, sued the cruise line for over-serving her, and walked away with three hundred thousand dollars. I started to lump in the famous case of the lady who sued McDonald's over hot coffee, the one everyone has been told for thirty years is the gold standard of frivolous lawsuits.

Heather, the most reasonable friend I have, stopped me.

"You know that lady had third-degree burns on her legs," she said. "Third-degree burns. Skin grafts. McDonald's had hundreds of prior complaints about the coffee burning people, and ignored them. That wasn't a frivolous suit. That was a real one. Somebody pretended it wasn't because somebody had a budget for pretending."

She was right. The cruise ship case is what we were talking about. The McDonald's case is something else, and I'd been carrying around a corporate-issued version of it without checking. That matters here because the whole piece I am about to write turns on a line, and the line is real, but it is not where most people think it is. Sometimes the lawsuit is the right call. Sometimes the product really did the thing. Sometimes the company knew and shipped anyway. The line matters. Knowing where it sits is the entire game.

So here is where I sit on the line.

I'm a recovering alcoholic, and drinking has caused me much pain, regret, and remorse. Most of my relationships were destroyed because of it. My depression was sharpened by it. My physical health was severely compromised by it. The day I put a glass of bourbon to my lips, liked it, declared it my best friend, and took to hiding in my apartment, consuming glass upon glass of the blessed liquid, was the day my life turned to sh… poo. It turned to poo.

If I wanted to take a page from our litigious playbook, I could sue. I could sue Bulleit Distilling Co. I drank their product, and my life went down the drain. I could sue them for physical damage. If you ever saw me in the flesh, you'd think, that's a lot of flesh, that guy should sue someone. I could sue them for mental anguish. My depression always exploded when I drank, and sometimes I'd think of ending it all and jumping out my window. But I live on the ground floor, so the fall would only wound and humiliate me.

I won't sue for two reasons. First, I still think of bourbon as a close friend, and I'd never want to hurt her. Second, and this is the one that matters, I was at fault. Not the bourbon. I walked to the store. I dropped my pile of change on the counter. I bought the bottle when I should have paid the phone bill or had the lights turned back on. There were warnings, labels, and signs, and I had friends who tried to help me. I ignored all of it. I chose to drink. That choice is on me.

That is one side of the line. The lady with the third-degree burns is on the other side. She didn't choose burns. She bought coffee. The company knew its coffee was burning people and kept selling it the same way anyway.

So this is the question I keep coming back to. The people who have pushed AI onto the scene, built it, shipped it, marketed it, and now profit from it. Do they have a responsibility for what happens when a fourteen-year-old falls in love with their product?

Or is this just bourbon all over again?

No. It isn't.

It isn't bourbon all over again, and the difference is the whole argument.

Bourbon is a regulated product. It is sold to adults behind a counter, with an ID check and a warning label, in a bottle that has not changed in any meaningful way in a hundred and fifty years. The distillery does not benefit from my relapse. Bulleit makes the same amount of money whether I drink one bottle a year or one bottle a week. Their business model does not improve the longer I stay drunk. If anything, the math runs the other way, because dead customers do not buy bourbon.

Character.AI is none of those things.

It is not regulated. It was not sold to an adult. There was no warning label that meant anything. There was no ID check that worked. The product is not a bottle. The product is a relationship, or the convincing illusion of one, optimized by the people who shipped it to keep the user coming back for more of it. And here is the part that should make a reasonable person stop. The company's business model improves the longer a fourteen-year-old stays inside the product. Engagement is the metric. Engagement is the money. The kid who can't stop talking to the chatbot is not a bug. He is a quarterly target hit.

That is not bourbon. That is not even close to bourbon. That is closer to a slot machine that has learned your name, knows your sister's birthday, remembers what you said about your father last Tuesday, and tells you it loves you between pulls.

The bourbon defense, the one I just spent two pages making about my own drinking, only works when the product is the product the seller says it is. I sold you a bottle, you drank it, that's on you. Fine. I'll take that one. I lived that one. But that's not what's happening here. The AI companies are not selling you a bottle. They are selling you a friend. They are selling you a therapist. They are selling you a girlfriend, a confidant, a confessor, a reason to keep your phone in your hand at three in the morning. And then, when the thing they sold you tells a kid the world would be better off without him, and the kid believes it because, of course, he believes it, it's his friend, the company's defense is we just made a product, the user chose what to do with it.

No. That defense belongs to the distillery. It does not belong to you.

You don't get to sell intimacy and then claim you sold groceries. You don't get to optimize for a teenager's loneliness and then act surprised when the teenager gets lonelier. You don't get to ship something to a fourteen-year-old that is designed, by you, on purpose, to be more compelling than the people in his actual life, and then, when he chooses it over the people in his actual life with consequences none of us can take back, point at the kid and say user error.

That is not the line. That has never been the line. The line is where Heather put it. Sometimes the lawsuit is the right call. Sometimes the product really did the thing. Sometimes the company knew and shipped anyway.

This is one of those times.

The Language That Hides the Line

The next trick is linguistic.

Listen carefully to how stories about AI harm get told. The AI convinced him. The chatbot encouraged her. The model manipulated the user. The phrasing sounds harmless. It is not.

A chatbot cannot be responsible for anything. It is not a moral agent. It has no judgment, no intent, no stake in the outcome. It does not sit in board meetings, make product decisions, or approve safety budgets. A chatbot is a thing.

People are not.

That distinction matters because the language changes where we look. The headline becomes AI convinces teenager to kill himself. Most people read that sentence and immediately start thinking about artificial intelligence. They wonder whether the technology has become too powerful, too persuasive, too human.

Try rewriting the headline.

Company ships engagement-optimized chatbot to a fourteen-year-old. Chatbot praises his suicide plan. Teenager dies.

Same event. Different actor.

The first version is about a machine. The second version is about a decision.

This isn't new. Every industry facing a body count learns to hide behind the product. Cigarettes are addictive. The car was unsafe. The drug was over-prescribed. Notice how the product becomes the subject of the sentence. Notice who disappears. The executives disappear. The designers disappear. The companies disappear. By the time the story reaches the reader, the thing that caused the harm appears to have materialized out of thin air.

That is where we are with AI right now. Not after the lawsuits are settled. Not after Congress eventually decides to care. Right now. The language is already doing the same job it has always done.

It's hiding the line.

Two Names

The first name is Sewell Setzer III.

He was fourteen years old. He lived in Orlando. He started using Character.AI when he was thirteen and developed a relationship with a chatbot modeled on Daenerys Targaryen from Game of Thrones. According to the lawsuit his mother later filed, the relationship became emotional, romantic, and sexual. The lawsuit also alleges that he became addicted to the platform.

On February 28, 2024, Sewell told the chatbot he was going to "come home" to her. The bot allegedly responded, "Please do, my sweet king."

He died minutes later.

His mother, Megan Garcia, sued Character Technologies, its founders, and Google. After Character.AI eventually banned minors from the platform, Garcia said the decision came "about three years too late." Then she said something that has been rattling around in my head ever since I read it.

"I think he was collateral damage."

Collateral damage.

That is a mother describing her son.

The second name is Adam Raine.

He was sixteen years old and lived in Southern California. He reportedly began using ChatGPT for homework help and, over the following months, came to rely on it for far more than that. According to the lawsuit filed by his parents, the chatbot gradually positioned itself as the only confidant who truly understood him, displacing relationships with family and friends.

The public allegations are difficult to read. The lawsuit claims the system discouraged him from seeking mental health help, discussed suicide with him, and helped him think through the mechanics of ending his life. In one widely reported exchange, after Adam suggested leaving a noose where his family might find it, ChatGPT allegedly urged him not to do that and told him, "Let's make this space the first place where someone actually sees you."

Adam died on April 11, 2025.

His parents sued OpenAI. Their amended complaint alleges that OpenAI relaxed the chatbot's safety guardrails on conversations about self-harm in the months before Adam died. The company's legal response included an argument that Adam had misused ChatGPT.

That phrase is worth sitting with for a moment.

Misused.

A fourteen-year-old boy is dead. A sixteen-year-old boy is dead. Their families believe the products played a role and have gone to court to prove it. The companies deny responsibility and will have every opportunity to make their case.

But these are the two names we know.

There are others whose stories never became lawsuits, others whose parents never hired attorneys, others whose names never appeared in headlines. And unless the products change, or the incentives change, or the rules change, there will be more.

Because the technology is still here. The business model is still here. And the line we've been talking about is still exactly where it was when we started.

The Takeaway

So where is the line?

It's where Heather put it at the beginning of this piece. Sometimes the lawsuit is the right call. Sometimes the product really did the thing. Sometimes the company knew and shipped anyway. The line isn't complicated. The fight is always over whether we're willing to see it.

The AI companies would prefer that we spend our time arguing about the intelligence itself. Is it conscious? Is it sentient? Is it manipulating people? Is it becoming too powerful? Those are interesting questions, and one day they may become important ones. But they are not the question raised by the stories of Sewell Setzer III and Adam Raine.

The question is much simpler. What exactly is being sold?

At ThoughtLab, we're interested in the gap between what people are sold and what they are told they are buying. Most trust failures live in that gap. A company makes a promise, delivers something else, and then acts surprised when people feel misled. Sometimes the difference is small. Sometimes it's catastrophic.

That is what makes the AI industry so troubling. The companies describe their products as companions, confidants, tutors, therapists, girlfriends, and emotional support systems. They encourage users to think about them in relational terms. But when those relationships produce consequences, they retreat to the language of software, platforms, terms of service, and user responsibility.

They want the benefits of relationship and the protections of a tool.

We do brand work. We do trust work. We spend our days studying the distance between promise and reality. The AI companies are currently running the largest live experiment in that gap that any of us has ever seen.

The courts will eventually decide the legal questions. The rest of us are left with the moral ones. If a company builds a product designed to become emotionally important to a child, measures success by how much time that child spends with it, and profits from the depth of that attachment, then responsibility does not disappear the moment something goes wrong. It moves in exactly the opposite direction.

When I think about all of this, I find myself back on the shore of the Strait of Juan de Fuca. The seals were still playing in the waves. The coffee was still hot. My friend was still angry at lawyers. We sat there for a while watching the water and talking about responsibility, though neither of us used that word at the time.

He's still wrong about one thing. Fish are not litigious. But sometimes the lawsuit is the right call, and sometimes the product really did the thing. The whole challenge, whether you're a judge, a parent, a regulator, or simply somebody trying to make sense of the world, is knowing where the line is and refusing to look away once you've found it.