Associate Foundations · 15% of the exam

governance, risk and responsible use

This exam domain is about a seventh of the exam. It covers the judgement around the work rather than the work itself: which tasks should go to Claude at all, what data may reach it, how an organisation's rules for AI are owned and followed, and what an output means for the people it reaches or describes.

Course example

Every domain follows Wrenfield Housing, the course's own example rather than an exam scenario. Wrenfield is a regional, not-for-profit housing provider with about 300 staff. Its operations, communications and project teams use Claude on the organisation's Team plan. Its AI policy has two rules: tenants' personal data stays out of anything sent to Claude unless it is anonymised first, and a person checks anything that goes to tenants, the board or the regulator.

Domain 6 puts those two rules to work. The operations team lead's quarter list, where every task was marked for Claude, is sorted into tasks that fit, tasks that need a named checker and tasks that stay with people (6.1). The income team takes the names out of an arrears export before Claude sees it (6.2). A new starter's slip with a personal account is reported the same day and closes a gap for everyone (6.3). A hardship fund FAQ with every fact right still shuts some tenants out until the team checks who it leaves behind (6.4).

"Claude can write it" is not the testkeep the task, take the people outask first, report earlyaccurate is not the same as fair
1,480 rowsof arrears data analysed with no tenant named in them. The key that turns codes back into accounts never left the income team.

Start hereThe words used throughout this domain

use case
A task you might give Claude. An appropriate one is allowed, uses data that may be shared, leaves decisions about a person with people, and has someone who can check it (4.1, 6.1).
personal data
Any information about a person who can be identified, directly or by putting details together. Anonymising changes it so that no one can be (6.2).
Usage Policy
Anthropic's rules for using its products. It applies to everyone, and an organisation's own AI policy adds to it but cannot allow what it forbids (2.4, 6.3).
governance
The processes around an AI policy: who owns it, who approves new uses, how slips are recorded and when it is reviewed (6.3).
bias
A pattern in an output that favours some groups of people over others, often through an assumption nobody stated (2.2, 6.4).
prompt and Project
A prompt is the message you send Claude, with everything you give it (1.1). A Project is a workspace whose chats share the same files and standing guidance (5.1).

task statements

flashcards

0 / 12 known
Card 1 of 12