Data & Databases
Training data leakage
AI learning or being tested with information it would not have when doing the real task.
Example
A churn AI system accidentally uses a field recorded after a customer cancels.
Why people use it
It helps explain why a promising test score may not survive real use.
What you'll hear
“That field wouldn't exist when the prediction is made.”
What this means for you
Check when each feature becomes available and keep test information isolated.
Can you control it?
No
No direct control. This describes a wider issue, concept or result rather than something you can simply switch on or off in a tool.
Common questions
- Does leakage here mean that private information was stolen?
- No. Here it means the learning or test had an unfair information advantage, making results look better than they should.
- Can the problem happen accidentally?
- Yes. A field recorded later or an unnoticed duplicate can give away information the system should not have.
- Does removing the obvious field always fix it?
- No. Other fields may indirectly reveal the same unavailable information.