Embed in a route
<FigureFromSource sourcePath="content/publications/figures/products/adversarial-range.dsl.md" figureId="RANGE-02" />
RANGE-02
Evaluation-to-Regression Lifecycle
An observed failure becomes durable security capability when it is converted into a replayable regression fixture.
governed lifecycle
Evaluation lifecycle
- 1Define the scenarioRecord target assumptions, preconditions, allowed actions, and expected evidence.
- 2Execute under controlRun within explicit authorization, data, rate, and environment limits.
- 3Observe the resultCapture model, retrieval, tool, agent, and external behavior.
- 4Reproduce and challengeRepeat the behavior and test alternative explanations.
- 5Create the regression fixturePreserve the minimum replayable inputs, assertions, and evidence hooks.
- 6Rerun after changeDetermine whether the failure is closed, residual, rejected, or still reproducible.
Retest loop returns to Execute under control
What did the rerun prove?
- Closed
- Residual
- Still reproducible
- Inconclusive
About this figure
Evaluation-to-Regression Lifecycle describes a governed security workflow in which an observed failure is not treated as a one-off anomaly but as durable evidence. A scenario is defined, executed under controlled conditions, and its result is observed and challenged through reproduction. Once the failure is confirmed, it is converted into a replayable regression fixture that captures the behavior precisely enough to be rerun after subsequent changes. The lifecycle closes the loop between evaluation and enforcement, turning adversarial findings into lasting checks that preserve the capability to detect the same failure class in future revisions.
Citation
Evaluation-to-Regression Lifecycle (RANGE-02). AI Security LLC Figure Library. https://aisecurity.llc/publication-dsl/figures/RANGE-02