Microsoft PCM: Попытка откупиться от авторов или новый рынок на миллиарды?

The era of free lunch for neural networks is officially coming to an end. For a long time, AI developers treated the open web as an endless buffet, downloading everything — from Reddit comments to Pulitzer Prize investigations — without asking and certainly without paying. But when the number of lawsuits began to grow exponentially, Microsoft decided to change tactics. Instead of endlessly feeding an army of lawyers, the corporation is building Publisher Content Marketplace (PCM). This is a digital storefront where any author or publisher can list their content for sale, and model developers can buy a license to train.

This step did not happen in a vacuum. The industry faced a problem called the "data crisis." We are rapidly approaching a moment when quality texts written by humans will simply run out on the internet. Training new models on data generated by previous versions of AI is a path to disaster and intellectual degeneration of neural networks. Microsoft understands that for a hypothetical GPT-5 or its successors, fresh, verified, and deep knowledge is needed. PCM is an attempt to centralize the supply chain of this "fuel" before competitors like Google or Apple divide it through exclusive contracts.

The irony of the situation lies in Microsoft's role. The company is not just acting as a buyer; it is creating infrastructure for the entire market. This is a classic "selling shovels during the gold rush" strategy, where the shovels are legally clean tokens for training. By creating such a marketplace, Microsoft appoints itself the supreme arbiter in relations between creators and technology. The message to publishers is crystal clear: no need to sue, just name your price in our application. This is an elegant way to neutralize legal threats while taking a commission on every transaction.

For publishers and authors, this offer looks like a double-edged sword. On one hand, real money will finally flow into the industry for content that was previously simply stolen. On the other hand, copyright holders are essentially selling seeds for their own disappearance. If a neural network learns to write exactly like a specific journalist, or analyze markets like a top agency, why would a user need the original source? Microsoft's platform may give media a short-term financial injection, but it accelerates a future where AI becomes the only interface for getting any information.

The barrier to entry into the top league of AI development is now getting even higher. If OpenAI, Google, and Meta can afford to spend hundreds of millions of dollars on licenses through such platforms, small garage startups don't have such budgets. The marketplace effectively puts quality data behind a high fence, with entry only by gold card. We are witnessing a transition from an era of open innovation to an age of corporate data repositories. The question is only what share of this pie Microsoft will leave to those who actually create the content.

The bottom line: The "Wild West" of AI training is closing for retraining. Microsoft is building a cash register where everyone will line up — both those who want to survive by selling their archives and those who desperately need these archives for model survival. Will there be any place in this world for those who don't have an extra billion for a subscription?