What AI training buyers pay for, how deals are priced and structured, and what an owner should check before signing anything. Written by Data Truffle. New posts most nights.
Five things move the number more than anything else: years of history, volume, outcome labels, modality, and rights. How each one changes what a buyer will pay, and what you can do about it before going to market.
Read →Inbound offers for operational data are landing in owners' inboxes. Most are legitimate. Some are not. Six signals that separate a real buyer or agent from a bad deal, and the question that exposes each one.
Read →Before a buyer trains on your records, the people in them are removed. Here is what that means in practice, what a de-identified ticket looks like next to the original, and what belongs in the contract.
Read →One buyer names one number. Several buyers reveal what the records are worth. A widely reported 2026 bankruptcy auction shows the gap, and the same logic applies to a company that is not for sale.
Read →A manifest is an inventory, not a dataset: what records exist, in which systems, over how many years, with what rights. It is the page buyers price from, and the reason you never send data to get a number.
Read →Models have read the public internet. What they have not seen is how a real company runs. That gap is why buyers now license operational records, and why yours may be worth more than you think.
Read →Measured from a complete data manifest. A first offer is a process target, not a guarantee; buyers decide eligibility, price, and terms.