Collecting only what you actually need. It runs against the instinct to gather everything in case it becomes useful, and it is both a legal requirement in many places and a practical risk reduction: data you never collected cannot leak, cannot be subpoenaed and cannot be misused.
Technical view
A principle in data protection law requiring data to be adequate, relevant and limited to what is necessary for a specified purpose. In machine learning it conflicts with the fact that feature usefulness is often discovered rather than predicted, and the resolution is purpose-specific collection with documented justification rather than blanket retention. Feature importance analysis after the fact can support removing inputs that contribute little.
Related
Part of AI Terminology Index, section: Data.