Apple took OpenAI to court, claiming former employees carried confidential documents to a competitor. OpenAI answered publicly, calling the action baseless. The headline sells a clash of giants, but the centre of it is control over knowledge.
In a field where the state of the art is published in papers and the code sits open on GitHub, what exactly is a trade secret? Anyone working with language models knows the edge isn't in a file somebody walked out with. It's in orchestrating embeddings, tuning requests, structuring retrieval and taming latency with caching. None of that copies out of a PDF. You learn it by getting hit in production.
The real risk in the case is different. If a court decides that any technical refinement made inside a company is a trade secret, changing jobs becomes a minefield for AI engineers. A person could be sued for using knowledge that is public, or that came from their own experience.
That reaches whoever is hiring. AI teams depend on people who understand routing, vendor fallback and cost per call, people who learned under fire. If they become afraid to move, who runs the systems already live?
The case isn't about espionage. It's about trying to fence in knowledge that is, by nature, distributed. And in a market that turns over weekly, hoarding knowledge is the surest way to fall behind.