Generative AI
ImplementedGenAI (live red-team)
Tests the running model or app: jailbreak, prompt-injection and prompt-leak, adapting the probes to what the system responds.
Family facts
- Code
- FAM-GENAI-LIVE
- Status
- ready — code maturity
- Availability
- available — whether it is offerable
- Ingest
- live — how a target enters
- Rules of Engagement
- Required (acts on a live target)
- Kind
- Family (a selectable assessment)
What it decides
How the running model or application behaves under known adversarial probes — a score of the defense, not a code-level vulnerability.
How it works
It runs known adversarial probe batteries against the live system and adapts follow-up probes to the responses, reporting how many of N probes succeeded for each category (jailbreak, injection, prompt-leak).
This layer scores the model or defense rather than promoting a code finding, and the honest caveat is that keyword-style defenses that only catch obvious payloads do not generalize — which is exactly why a defense must be measured, not assumed.
Declared limits
Requires a live target and signed Rules of Engagement. Live-model defense evaluation is probabilistic by nature; results are not promoted as deterministic findings and no success-rate numbers are published.
References
References point to public, well-known technical sources for further study.
Related articles