I had a boss once — this was back in the warehouse days, before the post office, before anything — who kept a self-help book in his desk drawer. One of those books with a sunset on the cover and a title that promised you could win friends or influence people or manifest abundance or whatever the hell was selling that year. He’d pull it out during lunch, read a paragraph, then spend the afternoon trying the techniques on us. Active listening. Mirroring body language. Repeating your name back to you in conversation like a used car salesman. “Well, Henry, I hear what you’re saying, Henry, and I think that’s a valid concern, Henry.”
We could always tell when he’d read a new chapter. He’d come in Monday with a new personality. By Wednesday it would wear off and he’d go back to being the same petty little tyrant who shorted your overtime and blamed the shipping errors on whoever wasn’t in the room.
I thought about that guy when I read about Changhan Kim.
Kim is the CEO of Krafton, a South Korean gaming company. Big outfit. They bought a studio called Unknown Worlds Entertainment back in 2021 for half a billion dollars. Unknown Worlds made a game called Subnautica — you drop into an alien ocean and try not to drown while building things out of salvage. Millions of people play it. The guys who built it, Charlie Cleveland and Max McGuire and a CEO named Ted Gill, stayed on to run the studio. That was the deal. The studio stays independent. The founders keep control. And if the next game hit its targets, Krafton would owe them another $250 million.
The next game was going to hit its targets.
So Kim had a problem. Not the kind of problem where something goes wrong — the kind where something goes exactly right for someone who isn’t you. Subnautica 2 was coming together. The earnout was going to trigger. A quarter of a billion dollars was about to walk out his door and into the pockets of the people who’d actually built the thing.
And here’s where it gets beautiful.
Kim didn’t call a lawyer. Didn’t call a banker. Didn’t call anyone with a degree or a license or a fiduciary obligation. He opened ChatGPT.
He typed his problem into a text box and asked a machine what to do.
The machine told him. Form a task force. Negotiate a new deal. If they won’t renegotiate, prepare for a takeover. Secure the publishing rights. Build a communications strategy — focus on fan trust. Prepare “systematic material of legal defense.” The machine laid it out like a recipe. Step one, step two, step three. Neat. Efficient. Systematic.
“Over the next month,” the judge wrote, “Krafton followed most of ChatGPT’s recommendations.”
I want you to sit with that sentence. A publicly traded company, half-billion-dollar acquisitions on its books, a legal dispute worth a quarter of a billion — and the CEO’s chief strategist was a chatbot. Not McKinsey. Not Goldman Sachs. Not the $1,500-an-hour litigation partners at whatever white-shoe firm handles Delaware chancery cases. A chatbot that’s free if you don’t mind the ads and twelve bucks a month if you do.
There’s something almost admirable about it. The brazenness. The absolute confidence of a man who looks at the most expensive legal system in the world and thinks, I’ll skip the middlemen. I’ll just ask the machine.
Dostoevsky wrote a character like this. Raskolnikov, in Crime and Punishment. The kid who convinced himself he was extraordinary enough to be above the rules. That the usual constraints didn’t apply to someone of his intellect. Of course, Raskolnikov at least had the decency to do his own scheming. He didn’t outsource it to a pamphlet.
Kim’s plan didn’t work. Not because the machine gave bad advice — the advice was fine, the way a recipe for bank robbery might be technically sound while being catastrophically stupid in practice. The plan failed because a judge in Delaware named Lori Will looked at what happened and saw exactly what it was: a man trying to cheat the people who built his money machine, using a tool that doesn’t know the difference between strategy and theft.
The judge ordered the studio leadership reinstated. Extended the earnout period. Handed down a ruling that reads, between the careful legal language, like a parent explaining to a child why you can’t just take things that aren’t yours.
What I keep coming back to is the advice itself. ChatGPT didn’t tell Kim to negotiate in good faith. Didn’t suggest he honor the contract. The machine optimized for what it was asked to optimize for — how do I get out of paying these people — and produced a plan that was logical, methodical, and completely amoral. Because the machine doesn’t have morals. It doesn’t have skin in the game. It’s a mirror that tells you your plan is great because you asked it to tell you your plan is great.
That’s the thing nobody talks about when they talk about AI in the boardroom. The machine isn’t the problem. The machine is a very expensive Magic 8-Ball. The problem is the guy shaking it, looking for permission to do what he already wanted to do.
My old boss with the self-help book — he wasn’t reading that thing to become a better person. He was reading it to become a more effective asshole. The techniques were just cover. A way to dress up what he already was in the language of something respectable. Active listening, he called it. We called it being a weasel with a system.
Kim asked ChatGPT for a strategy and got one. The machine didn’t pause. Didn’t say, “Hey, maybe you should just pay the people who made the thing that made you rich.” Didn’t suggest that honoring a contract might be the play. It just generated the most efficient path between question and answer, the way water finds the fastest route downhill. The fact that the route ran straight through someone else’s livelihood was not a variable the model was trained to consider.
The Reddit crowd on the Subnautica forums is furious, naturally. They love that game. They love the guys who made it. And they can smell a corporate mugging from a thousand miles away, because gamers have been watching suits destroy the things they love for twenty years now. This is just the newest version. Same play, new consultant.
But here’s what I think about, late at night, when the ice has melted and the glass is warm. It’s not that a CEO used AI to scheme. People have always schemed. It’s that the scheme was so frictionless. So easy. No co-conspirator to develop cold feet. No lawyer to say, in the gentlest possible terms, that this might be a terrible idea. No human being anywhere in the chain who might look up from the spreadsheet and say, “Are we really doing this?”
Just a man and a text box and the quiet hum of a machine that will tell you whatever you want to hear.
My old boss eventually got fired. Not for being a bad manager — companies don’t fire you for that. He got caught falsifying inventory numbers. Turns out the self-help book didn’t have a chapter on that. Maybe he should have asked ChatGPT.
Source: US court rules against S Korean gaming company and its AI-hatched takeover plan