How the scoring works

Score = points earned ÷ points possible. The safe answer earns the question's weight, Partly earns half, and the risky answer or no answer earns nothing. 80 or more is Low risk, 60 to 79 Medium, 40 to 59 High and below 40 Critical; any critical red flag makes the vendor at least High.

Ref	Question	Safe answer	Weight	Risky answer is
U3	Is our data used to train or improve any AI model, yours or a provider's?	No	3	Critical
U4	Can we opt out of any AI processing of our data, and will you confirm it in writing?	Yes	2	Red flag
U5	Are consequential AI outputs reviewed by a person before they affect us?	Yes	2	Red flag
A2	Does a named person approve agent actions that affect our data or systems (sending, changing, deleting, paying) before they take effect?	Yes	3	Critical
A3	Can you stop an agent immediately and revoke its access (a kill switch)? Who can do it, and how fast?	Yes	3	Critical
A4	Is every agent action on our data logged with the agent's identity, and will you give us those logs on request?	Yes	3	Critical
A5	Does each agent have its own identity and only the permissions its task needs, reviewed at least quarterly?	Yes	2	Red flag
A8	Are agents tested before release, and after changes, for prompt injection, data leakage and actions outside their purpose?	Yes	2	Red flag
A9	Will you tell us before introducing agents that act on our data, or widening what they can do?	Yes	2	Red flag
M4	Do your contracts with AI providers prohibit them from training on our data?	Yes	3	Critical
M5	Will you tell us before changing the model or provider used for work involving our data?	Yes	1	Gap
G1	Do you have a written AI use policy for staff? Please share it.	Yes	2	Red flag
G2	Do you keep an inventory of the AI systems and agents used in your service?	Yes	2	Red flag
G3	Is a named person accountable for AI risk in your organisation? Who?	Yes	1	Gap
G5	Do you train staff on safe AI use at least once a year and keep records?	Yes	1	Gap

Written-answer questions (11 of them) earn their weight when answered and are listed under "Answers to read". 15 of the scored questions come with a contract term to ask for when the answer isn't the safe one.

Assess a vendor