AI Bill of Materials
A parts list for an AI model, tying it back to the data, licences and training runs that produced it.
An AI Bill of Materials, sometimes shortened to ABOM, is a list of what went into a model. The idea is borrowed from manufacturing, where a bill of materials tells you every component in a product, and from software, where a software bill of materials tells you every library inside an application. Applied to AI, it records the sources of the training data, the licences claimed over that data, who contributed it and which training runs produced the finished model.
Almost no model available today ships with one, which is why the question “what was this trained on” usually has no answer. The concept matters most for two groups: companies that need to know whether they are legally allowed to use a model’s output, and regulators who would like the AI industry to be as inspectable as the open source software it is built on. OpenWALDO is the most visible attempt so far to make it practical.