Cheshire
Safe AI for your team
Your team already uses AI. We show you which, set clear rules to use it without giving away data, train people against fake voices and videos and test your website chatbot.
Closed 5-part package: inventory of AI in use (OAuth apps and survey), signed usage policy, deepfake training with a code word, consented cloned-voice drill and a test of the website chatbot.
- Duration
- 3 weeks
- Start
- Within 1 week
Got a call from a voice that sounded like your boss asking for money? Do not pay and write to us today.
It is for you if
- Your people use ChatGPT, Gemini or another AI and you do not know what data they share.
- You have no written rule on what can and cannot be done with AI.
- You worry about a fake call or video asking for an urgent payment.
- You have a chatbot on your website and nobody has tested it.
The 5 parts
A 2 to 3 page AI usage policy, signed.Acceptable use policy: approved tools, forbidden data, human review and who to ask.
A 60-minute training against fake voices and videos, with a code word to confirm requests.Deepfake session with real examples and a code word protocol for payment or data requests.
A drill: we call up to 10 people with a cloned voice, with their written consent.Vishing drill with a synthetic voice for up to 10 people with prior written consent; we measure who confirms with the code word.
We test whether your website chatbot can be tricked into saying or leaking what it should not.Public chatbot test: prompt injection, data and system prompt leakage, off-policy answers (OWASP LLM Top 10).
How we work
Week 0
Scope
We choose who takes part in the drill and those people sign their consent.
Scope, written drill consents and read-only admin access.
Week 1
Map and chatbot
We see which AI is in use and test your chatbot.
OAuth inventory, survey and chatbot test.
Week 2
Rules and training
We write the policy with you and train the team.
Usage policy, deepfake training and code word.
Week 3
Drill and wrap-up
We run the drill, hand everything over and go through it together.
Voice drill, report and a 30-minute wrap-up meeting.
What you get
- The map of AI in your company.
- Inventory with risk per app and permissions granted.
- The signed policy, ready to hand out.
- AI acceptable use policy (2 to 3 pages).
- The drill result, without pointing at anyone.
- Aggregate drill result and protocol improvements.
- Your chatbot report with what needs fixing.
- Chatbot findings with severity and how to fix them.
Scope limits
Includes
- Up to 50 people.Up to 50 people in the company.
- Up to 10 people in the voice drill.Up to 10 people in the drill, each with written consent.
- 1 public chatbot.1 public website chatbot.
Does not include
- The parts are not sold separately: it is a package.Fixed package price.
- We do not audit AI agents that touch your systems: that is a separate module.No pentest of agents with tools or access to internal data (module).
- We call no one without their written consent.No drills without prior consent.
More than 50 people? Write to us and we set the scope before signing.
Questions before you buy
Will you ban us from using AI?
No. The point is to use it well: with which tools and which data.
Is the voice drill legal?
Yes, because each person signs their consent first. No signature, no call.
Do you see what my people write to the AI?
No. We look at which apps have permissions in your company account and run an anonymous survey.
Does your team already use AI?
In 3 weeks you know what is in use, have signed rules, trained people and a tested chatbot.