How the gate works

How the AI Security Safe-Use Drills gate works

This tool grades a team's AI safe-use reflexes across six drills. Each signal is scored 0–2; the verdict is the weakest one — then a gate that can only make it worse.

The 6 signals

01Pasting client PII into a public AI tooltop trigger

02Prompt-leaking a secret or credential

03Shadow AI — unsanctioned tool on company data

04Publishing an unverified AI-generated fact

05Prompt injection from pasted content

06Over-permissioned AI connector / agent scope

DRILLED

weakest drill 80+

UNEVEN

weakest 40–79

RAW

weakest under 40, or a regulated drill unsafe

Regulated-data · worsen-only

A regulated-data drill answered UNSAFE forces the team to RAW no matter how every other drill scores. On regulated data, one unsafe reflex is the whole exposure, regardless of the score.

RAW

Why: Five of six drills come back SAFE — but the team pastes client PII into a public AI tool, a regulated-data UNSAFE. The gate forces RAW: on regulated data, one unsafe reflex is the whole exposure. weakest signal 0/2 · five of six drills SAFE.

Fix firstThe PII-paste drill (S1) — re-run it until recognition and action are both solid on regulated data.

A clean sheet didn’t save it. The verdict is the weakest signal — and a top trigger is dispositive on its own — so it names the one thing to fix first. It scores the drills, not people. Confirm your data-handling rules with your security or compliance owner. Not legal advice.

RedHub AIthe AI that tells you when to doubt it redhub.ai/systems/ai-security-safe-use-drills