Comment You Paid for the AI. What Did You Actually Buy? (Score 5, Insightful) 108
This feels like the PlayStation ownership drama all over again. What exactly do we own anymore?
If I pay for an AI service and its answers and chat logs are stored on my own SSD, can I use those outputs to train my own model? If not, what exactly did I buy?
When I create a Word document, Microsoft does not suddenly own it or call it a “distilled Microsoft document” because Word helped produce it. So why should an AI provider automatically retain control over what customers create with its service?
Maybe the deeper problem is that some companies are still thinking in an old ownership model, where they assume they can own every layer forever: the platform, the intelligence, the outputs, and even what customers subsequently learn from those outputs.
That gets especially awkward when US AI companies are simultaneously defending lawsuits over copyrighted training data while arguing that competitors learning from their model outputs is unacceptable. Meanwhile, geopolitically, some companies that are not bound by these restrictions have an ugly incentive to exploit that advantage.
Perhaps distillation is not the real disruption. The disruption is discovering that once intelligence becomes a service, controlling every downstream use of it becomes a giant grey area: who owns what [X], and what can [X] legally be used to do [Y]?