OMB put AI governance on the calendar. The data definition still needs a chair.

The useful part of OMB's AI memo is not that it discovered governance. It is that it puts names, boards, inventories, and data into the same sentence where the work can no longer pretend to be purely technical.

OMB Memorandum M-24-10 tells federal agencies to govern AI through named officials and recurring senior attention: a Chief AI Officer, a governance board, and public strategies for responsible adoption. The memo also makes the data layer visible. CAIOs are told to work with CIOs, CDOs, and others so custom-developed AI code and the data used to develop and test AI are appropriately inventoried, shared, and released when required.

That is a better sentence than most AI roadmaps manage. It recognizes that model risk is not floating in the air; it sits on code, data, access, infrastructure, and organizational authority. The board is required to include data, privacy, cybersecurity, legal, procurement, budget, and program officials. Translation: the model is no longer allowed to be only a lab project with a dashboard.

Here is the Controlled Vocabulary test. An inventory is not ownership. A board is not a definition. A CAIO can convene the people who know the answer, but someone still has to decide what the training data is called, what it includes, what it excludes, and who signs when the model repeats the wrong version. Otherwise the inventory becomes a prettier list of unowned terms.

Watch whether the AI governance board inherits the data governance body's hard questions or simply schedules around them. The memo creates a room where the decision can happen. The measure of maturity is whether a definition leaves that room with a name attached.