The Odyssey you should avoid.
Every question your team types into an AI tool sends a little more out. Read what the contract says comes back.
By Amit Srivastava · First published on LinkedIn, 11 September 2026
Odysseus got home. He also lost every ship he sailed with. And every man.
A woman spent months building her own way of working inside ChatGPT, then went to court to protect it. The judge threw the case out. Typing it in had already given it away.
Picture the best thing your company knows.
Not the product. The way you work. How you set a price on a hard job. Which customers are worth chasing and which are not. The thing a new person takes two years to learn. The thing a rival would pay a lot of money to read.
Now picture somebody in your team pasting it into a chat box, because they wanted help writing it up neatly.
What the court said
Rebecca Trinidad built her own way of working while using ChatGPT. She believed it was hers. She went to a court in California to protect it.
In January, the court threw the case out. Not because her work had no value. Because she had typed it into ChatGPT. That meant she had handed it to OpenAI herself. A secret you give away is not a secret any more. The law only guards what you have kept.
Any leader should stop at that sentence. The protection did not fail. There was never any protection, from the moment the words went into the box.
Your way of working stopped being yours the moment you asked for help with it.
The long way home
Half the cinemas in the world spent this summer watching Odysseus try to get home. Read the poem again and you notice the cost. He set out with twelve ships. He came home on none of them. Every man who sailed with him died on the way.
He survived. Almost nothing he started with did.
That is the journey most companies are on with AI. The trouble is that it does not feel like a journey at all. It feels like a series of small helpful days.
The island of Lotus eaters
Odysseus lands on an island of Lotus eaters. His men try it. They simply stop wanting to go home. They forget what they came for. There is no villain in that scene.
The AI tool on your team's screen is a real gift. It does write the summary faster and in the right template. It does tidy the proposal. Nobody is being tricked. No single day goes badly.
Nobody watches what goes the other way. Every question carries a small piece of how your company thinks. Your prices. Your customer list. The argument you plan to use in a meeting next week. Your own people happily sending it out, one paste at a time.
And they believe they are being helpful.
Read your own contract
Most leaders think having a paid business account settles this. Often it does not. Not in the case of AI companies. A company called TermScout, working with Stanford's law school, compared AI contracts against normal software contracts.
In 92 out of every 100 AI contracts, the seller claims the right to use your data for more than just running the service. In normal software it is 63.
The words are docile. The seller may use your data to improve, build or enhance the product. That line was written years before anyone trained a model. Today it does cover training a model.
Only a 33% of AI contracts protect you if somebody sues over what the tool wrote. In normal software it is nearly 60%.
Twelve ships went out. None came back.
Your team types a question
How you set your prices
Your best customers
Two years of learning
The model
It goes in. It stays in.
Not yours.
What the contract says
Can use your data for more than the job
Protect you if somebody sues
Clearly promise to follow the law
The One way door
Here is the part with no fix. Your thinking does not sit in a folder you can delete. It spreads into the settings and weights inside the model.
Lawyers call this a one way door. You can delete a file. Nobody can pull one company's ideas back out of a trained model without pulling the whole model apart.
Now hear it from the other side
You might expect this warning from someone who campaigns about privacy. It came from the man who sells the tools.
In June, Satya Nadella the head of Microsoft wrote about what is at stake for companies. He was writing about how a company keeps learning, keeps its own ideas, stays different from everyone else.
His words were that AI models can learn the skills of people and companies. They turn it into something ordinary that anybody can buy.
Read that again. The head of Microsoft is saying it out loud.
Nadella is plain about where this ends. Nobody should want a world, he writes, where every company in every industry hands its value to a few models that eat everything they see.
What this is really about
Rebecca Trinidad lost one court case. One person, one set of methods.
Now think about a company of four thousand people. Each one asks a helpful question every day. Each question carries a small piece of how the place really works.
Every leader reading this has a view on which AI tools are allowed. Almost none have a view on what may be typed into them.
Those are two different questions.
Whether the tool is safe?
What your company is worth, then whether any of it will still be yours in five years?
Nobody here did anything wrong. Not Rebecca Trinidad. Not your team. The tool is useful, the contract is signed, the food is good.
Odysseus got his men off that island because he was the only one who had eaten the Lotus and still wanted to go home. Somebody has to stay awake enough to notice what the ships are losing.
