Run Binding refers to linking a specific AI run to a defined release state. It is the connection that makes each execution explainable in terms of the data conditions it ran under.
When a run finishes, only the answer remains. Which schema, which field definitions, which applied policy produced it disappears unless it is bound at the time. What an audit asks for is not the result but the conditions behind it, and without that link the result cannot be defended.
This differs in purpose from experiment tracking. Experiment tracking records hyperparameters and metrics so training runs can be compared. Run binding ties a production execution to a data state so it can be returned to under the same conditions. A bound run is reproduced through its release state, and if the state has moved, a diff shows what changed.