This is AI Frontiers. Today, the monopoly price you don't see. The monopoly problem in artificial intelligence has not disappeared. The extraction mechanism has changed. Access can stay cheap or free, while data, prediction, and switching costs quietly change who captures the value. Start with the textbook. A single price monopolist restricts output, charges more, and captures part of the surplus that would otherwise have stayed with buyers. Some surplus is transferred, some is destroyed outright. Now remove the price tag. A search engine can be free. A chatbot can be free. An assistant can arrive as one more feature inside software you already pay for. The temptation is to conclude that consumer surplus must be enormous. Sometimes it is. But a zero monetary price is not zero economic exchange. Users still give up behavioral information, attention, privacy, contractual rights, control over accumulated context, interoperability, and the practical ability to switch provider later. AI adds a second transaction. The user gets an answer. The system gets an interaction. A prompt reveals intent. A correction reveals what the system got wrong. A repeated sequence reveals a workflow. One correction is noise. Millions of corrections reveal where systems fail. The chain runs like this. Use generates information. Information improves prediction. A better product deepens dependence. Dependence shifts bargaining power. Then the export illusion. Suppose you can download all your files. What about conversation history, system instructions, agent state, embeddings, integration logic, permissions, and the behavior employees have learned to expect. If those cannot be reconstructed elsewhere, you can legally own your data and still be economically locked in. So apply the exit test. If I leave tomorrow, how much of the value I created here comes with me? Be precise about the slogan. If the product is free, you are the product is catchy but economically imprecise. A service at 20 euros a month that saves a professional 10 hours of work creates enormous consumer surplus. Consumer benefit does not disprove market power. Research by Bergamon, Brooks, and Morris shows that additional information and market segmentation can distribute consumer and producer surplus very differently. More data does not always hurt consumers. It changes what the seller can learn about demand and therefore what it can extract. The question is not whether the service is free. The question is whether you remain free to leave. That's AI Frontiers.