Data leakage
Also: test contamination
Data reaching a place it should not. In model tests, it can also mean test answers have reached the training data.
The fuller definition
In evaluation, data leakage means information unavailable in the intended real task improperly influences training or assessment. In security discussions, the phrase can also mean unauthorized disclosure.
- Held-out answer
Should remain unseen
- Leak
Answer reaches development
- Score
Misleadingly high
A simplified relationship, not a screenshot or a complete design.
A concrete example
A test answer accidentally included in a prompt inflates the measured score.
The distinction that matters
Keep evaluation contamination distinct from confidential-data disclosure; the remedies differ.
Related terms
Test set
Examples held back from training so you can check how the model does on things it has not seen.
Reliability & securityBenchmark
A standard test used to compare models. A high score on it does not mean the model is good at your task.
Reliability & securityConfidentiality
Keeping private material from people who may not see it. Check who can read, store, or share what you put into a tool.
Where it is taught
Test an AI workflow on work you can assessSource and further reading
Anthropic · Evaluating AI agentsThe definition and example are written for this library. The link gives background, official documentation or legal authority as labelled. Legal scope depends on the jurisdiction and current text; product features depend on release and settings.