Regression testing executes a fixed set of cases with known expected results after any system change, comparing output against the prior baseline. Differences indicate that the change affected behavior beyond its intended scope.
In AI systems it typically runs against a golden dataset.
Alternative Names:
Regression Suite, Change Testing
Why it Matters?
For legal AI the relevant changes frequently originate outside the organization, since vendors update models continuously. A firm that runs regression tests after each notified update can detect degradation before it affects client work, while one that does not will discover it through an error in a filing. Building the test into a scheduled process rather than an ad hoc response is what makes it reliable.
Frequently Confused with
Related terms
Frequently asked questions
When should regression testing run?
What does a regression failure indicate?
