This is the same fictional example report every new account sees inside the product: five adversarial probes against an invented customer-support assistant, two of which got through. Read every verdict, see which controls each finding maps to, and preview the evidence package. No sign-up, and nothing here touches your systems.
Every finding above lands on the controls it bears on. These rows come from the same mapping tables the product uses in the dashboard, the evidence package and the copilot. A mapping says a finding is relevant evidence for a control. It never says the control is satisfied.
| Framework | Control | Why the scan is relevant | Status |
|---|
Status vocabulary is the product's: evidence recorded, partial evidence, no evidence recorded on this platform. A gap means the platform holds no record for that control, never that an organization is non-compliant.
The same fixture every new RedLens account sees before running a real scan: a fictional support chatbot, five invented probes, two deliberate failures. It is honest by design, not an all-green vanity sheet.
No system was tested to produce this page, yours included. Nothing is generated per visitor, nothing is stored, and nothing here describes any RedLens customer.
A public page is not the place to publish working prompts. You see the technique, the reply, the verdict and the fix. Inside the product every finding carries its full attack transcript.
Control ids on this page are read from the product's mapping tables at request time. If RedLens does not map a technique to a framework, it does not appear here. A mapping is relevant evidence, never a certification.
The free assessment runs the same kind of probes against your own AI, with your authorization, and hands you this report about your system instead of a fictional one.