AI
The AI drafts. A person signs.
The checks and the assistants run fixed rules on your records. A language model answers questions: Ask, which uses Claude, from Anthropic. It reads through tools that change nothing, cites the records it read, and stays off until an admin switches it on. What it drafts waits for a person to approve.
Where a model is used, and where it is not
The checks
Each check is a fixed rule run on a record, and its result carries a fingerprint of what was checked.
The assistants
The assistants raise issues from fixed rules on your records, each with its evidence attached.
Ask
Answers questions in plain words from your records, using Claude, from Anthropic.
Ask, held in
With its sources
Each answer cites the records it read, and a number not found in them is pointed out.
Reads, and nothing else
It reads through fixed tools that cannot change anything, and email addresses are hidden from it.
Drafts wait for a person
A drafted email or action waits for a person to approve it. A marketing email to someone without consent is refused.
Off until switched on
For a live organisation it is off until its admin switches it on, and each person can ask thirty questions an hour.
Erased with the customer
Erasing a customer removes them from the questions, answers and drafts Ask kept.
Off in the demos
Ask is switched off in the free live demos.
Try this in a copy of your own, for two hours, with nothing to sign up for.
Open your own copyThe AI Act, in the audit
The audit asks about the AI an organisation uses: the tools and who uses them, the data put into them, a written policy, staff AI literacy, telling people when they deal with AI, and the uses the AI Act calls high-risk. These are the questions that raise a finding, and the fix each one leads to.
Is personal or confidential data ever pasted into AI tools?
Where it applies
SeriousPersonal or confidential data goes into AI tools.
The fixApproved tools only, on business accounts that do not train on your data, under a short written AI use policy.
Is there a written policy on which AI tools may be used, and how?
Where it applies
MediumThere is no AI use policy.
The fixA one-page AI use policy, signed by each person who uses AI.
Have the staff who use AI been trained on it?
Where it applies
MediumStaff who use AI have had no training on it, and the AI Act asks for measures that support their AI literacy.
The fixShort AI training for each user, recorded per person.
Are people told they are dealing with AI?
Where it applies
MediumCustomers are not told when they deal with AI.
The fixA clear notice wherever AI talks or writes to customers.
Is AI used to decide on hiring, credit, insurance, or access to services?
Where it applies
SeriousAI helps decide hiring, credit, insurance or access to services.
The fixChecked against the AI Act's high-risk list, with a person reviewing every decision and the review recorded.
EU AI Act · Regulation (EU) 2024/1689. Any organisation that uses AI systems has duties under it, from measures that support its staff's AI literacy to the rules for high-risk uses, which apply from 2 December 2027 for most.
Start here
See it on a business of your own.
Open your own copy in a few seconds and use it for two hours, with nothing to sign up for. When you want it on your own systems, we start with an audit.