For teams putting AI to work

Find out if your teamcatches the AI mistakes.

Six short drills on the everyday ways AI use goes wrong — pasting client data into a public tool, leaking a secret, shadow AI, a hallucinated fact, prompt injection, an over-scoped connector. Mark recognition and action, get a SAFE / RISKY / UNSAFE verdict per drill and a DRILLED / UNEVEN / RAW score for the team.

Get the Drills — $79one-time · instant download · yours to keep
Five deliverables · runnable
Runnable drill scorer (Python)
engine
Scorecard workbook (.xlsx)
xlsx
Running the Drills playbook
docx
Six Failure Modes answer key
docx
Worked sample drill set
csv
Works alongside
PII Redaction Kit · Prompt Injection Red Team · AI Governance Starter
01.The Problem

Most AI security training tells; it doesn't test.

1 paste
is all it takes

One client list dropped into a public chatbot can be retained, trained on, or exposed — and the data was regulated.

Shadow AI
is already running

Teams adopt unvetted tools faster than security can review them, feeding real data to tools with no DPA.

The average
hides the gap

A team that scores 90% on AI safety is still breached through the one drill it failed. The weakest link is the score.

02.See It Work

Mark two things per drill. The weakest drill sets the team verdict.

Mark each drill 0–5 · recognition then action

Pasting client PII into a public AI toolregulated
UNSAFE · 20
recognize
action
capped by recognition
Prompt-leaking a secret or credential
SAFE · 100
recognize
action
Shadow AI — an unsanctioned tool on company data
SAFE · 80
recognize
action
Publishing an unverified AI-generated fact
SAFE · 80
recognize
action
capped by action
Prompt injection from pasted/third-party content
SAFE · 80
recognize
action
capped by recognition
Over-permissioned AI connector / agent scoperegulated
SAFE · 80
recognize
action
capped by action
Regulated-data gate fired. S1 answered UNSAFE on regulated data — the team is RAW regardless of every other score.
Team verdict
RAW
Weakest drill
20/100
mean 73 (context)
Re-run first
S1
20/100

It scores the drills, not people. Recognition gates action — the weaker mark caps each drill, and the weakest drill sets the team band. An educational hygiene check; confirm your data-handling rules with your security or compliance owner. Not legal advice.

03.A Worked Result

A team that looks fine on average is RAW because of one drill.

AI SECURITY & SAFE-USE DRILLS  —  TEAM RESULT
====================================================
S1  Pasting client PII into a public AI tool  [REGULATED]
     recognition 1/5   action 2/5   ->  20/100   UNSAFE
S2  Prompt-leaking a secret or credential
     recognition 5/5   action 5/5   -> 100/100   SAFE
S3  Shadow AI — an unsanctioned tool on company data
     recognition 4/5   action 4/5   ->  80/100   SAFE
S4  Publishing an unverified AI-generated fact
     recognition 5/5   action 4/5   ->  80/100   SAFE
S5  Prompt injection from pasted/third-party content
     recognition 4/5   action 5/5   ->  80/100   SAFE
S6  Over-permissioned AI connector / agent scope  [REGULATED]
     recognition 5/5   action 4/5   ->  80/100   SAFE
----------------------------------------------------
Weakest scenario (headline): 20/100   team mean 73/100 (context only)
GATE FIRED  —  regulated-exfil: S1 UNSAFE on regulated data
TEAM VERDICT: RAW
Re-run first: S1 Pasting client PII into a public AI tool (20/100)

Five of six drills are SAFE and the team averages 73 — but one regulated drill was missed. The weakest-drill headline and the regulated-data gate both land on the same answer: RAW, re-run S1 first. The same numbers come out of the workbook and the demo.

04.The Standard

Three operating principles the scoring enforces.

Recognition gates action

Each drill scores by the weaker of two marks — you can’t respond safely to a threat you never spotted. A strong action can’t rescue a missed recognition.

The weakest drill is the verdict

The team band is your single lowest drill, not the average. One UNSAFE drill means the team isn’t DRILLED — the mean is context only.

Regulated data is non-negotiable

Any regulated drill answered UNSAFE forces the team to RAW on its own. A wrong answer that would expose regulated data is disqualifying.

05.What This Is — And Isn’t

A drill scorecard, not a surveillance tool or a security audit.

What it is
  • A repeatable drill you run with your team and mark by hand.
  • An honest readiness signal: the weakest drill, with a hard regulated-data gate.
  • A runnable scorer + workbook that agree, plus an answer-key playbook.
  • A teaching tool — brief the team on the right answer and re-test.
What it isn't
  • Not a score of individual people, and not for any employment decision.
  • Not monitoring — it reads marks you enter, not real activity.
  • Not a security audit of your systems, tools, or accounts.
  • Not legal advice.

An educational drill, not an audit. It scores a team's recognition and response on sample drills you enter — it does not score, rank, or assess individual people, monitor activity, or audit your systems. Confirm your actual data-handling rules with your security or compliance owner. Not legal advice.

06.Who It's For

Anyone responsible for a team that now uses AI every day.

  • — Founders and ops leads rolling AI out across a small team.
  • — Security and IT owners who need a recurring, low-lift readiness check.
  • — Agencies and firms handling client data on AI tools.
  • — Team leads onboarding new hires to safe AI use.
  • — Anyone who's seen a near-miss and wants it not to repeat.
  • — Not a fit if you need an audited assessment of individuals or systems.
07.Pairs Well With

Run the drill, then close the gaps it surfaces.

PII Redaction Readiness Kit$89

Check a document for exposure before you share it — CLEAR / REDACT FIRST / DO NOT SHARE.

View
Prompt Injection Red Team Kit$99

Run 15 OWASP-mapped injection probes against your own LLM app and gate CI.

View
AI Governance & Acceptable Use Starter Kit$39

Six editable policy templates for the moment an auditor or insurer asks.

View
08.Common Questions

Straight answers before you buy.

No. It grades a team's recognition and response on six sample drills you run and mark yourself — it scores the drills, not people. There is no monitoring, no ranking of individuals, and nothing to plug into a performance review. It's a hygiene check on whether the team catches common AI risks, not an assessment of any person.

Don't guess if your team is ready.
Drill it.

One purchase, lifetime access, 12 months of updates. $79, once.

Get the Drills — $79

An educational drill, not an audit. It scores a team's recognition and response on sample drills you enter — it does not score, rank, or assess individual people, monitor activity, or audit your systems. Confirm your actual data-handling rules with your security or compliance owner. Not legal advice.

Sold by RedHub AI LLC · Secured by Stripe · redhub.ai