Microsoft PCM: Redmond opens a "data supermarket" for AI training

Remember the good old days, when the internet looked like an endless free buffet for neural network developers? Those days are officially numbered. Microsoft has decided to lead a process many in the industry considered inevitable, and is building the Publisher Content Marketplace (PCM). Put simply, it is a kind of App Store, but instead of apps it sells the rights to use texts, news and analysis. The Redmond giant no longer wants to watch its partners and competitors drown in endless lawsuits from angry media groups.

The context matters more here than the technology. For the past two years, the AI industry has run on a "download first, apologize later" basis. OpenAI, Google and Microsoft itself swallowed petabytes of data from the open web without asking permission, let alone offering money.

The result was high-profile lawsuits from The New York Times and other major players. Microsoft PCM is an attempt to legalize the process before the courts finally shut off the data taps. Heavyweights such as Vox Media, Associated Press and Condé Nast are already taking part in developing the marketplace.

These companies are tired of fighting scraper bots and have decided that if their content will be used anyway, they should at least get paid for it.

What exactly has changed in the approach? The key word is "grounding". Modern LLMs often hallucinate, and to produce current facts they need access to fresh, reliable sources. PCM will let developers connect directly to media archives and receive verified information in real time. Publishers keep control: they set the terms of use themselves and see detailed reporting. Microsoft promises the transparency that was so lacking in the closed deals between OpenAI and individual media groups.

Why does this matter for the market right now? Microsoft is effectively building infrastructure for a new knowledge economy. If data used to be the "new oil" that anyone could extract for free on someone else's land, Microsoft is now putting a fence around that land and opening a cash register. That is great news for large publishers whose archives are worth millions, but an alarming signal for small AI startups. Only the giants will be able to afford licensed content, which will widen the gap between Big Tech and everyone else even further. We are watching the open internet gradually turn into a walled garden where bots must pay at the gate.

It must be said that Microsoft picked the perfect moment. The company needs to protect its Copilot from legal risk, and PCM solves that problem elegantly. Instead of negotiating with each publisher separately, it is creating a standard. If PCM becomes the industry standard, Microsoft will control not only the compute and the models but also the raw material those models run on. It is a strategic move that cements the company's dominance in the AI food chain for years to come.

The takeaway: Microsoft is turning data licensing into a scalable business, in effect admitting that the era of free web scraping is over. Are you ready for neural network training to become a privilege of the rich?