Skip to content

CPD

See a marked task

One task from Data security and governance for AI in finance, in full: what you're asked, how it's marked, an answer, and the feedback and mark it gets. Pass at 80 out of 100.

Consumer v business accounts: retention, training and where data is stored

The task

Kestrel Lane Distribution Ltd is a UK wholesaler with 160 suppliers. Since September the company has had a business plan for an AI assistant, under a company contract that excludes training. Staff have been told to use it for work, but there is no written policy yet.

On Tuesday, Tom Ashworth, the purchasing manager, wanted to decide which suppliers to renegotiate with. On his phone he opened the AI chat app he uses at home, on a personal paid plan, and pasted in the full supplier list from a spreadsheet. It had 140 rows with: supplier name, annual spend, payment terms, sort code and account number, contact name, email and mobile number, and a notes column. Some notes say things like 'owner going through divorce, may accept lower price'. About 30 of the suppliers are sole traders. The app gave him a ranked list, which he emailed to himself.

On Thursday morning Tom told the finance director, Aisha Bello. He does not know whether his account's setting allows his chats to be used for training, and he has not deleted the chat. Aisha has asked you, as the company's management accountant, for a note she can act on today.

  1. Explain the risks, separating data protection, commercial confidentiality and fraud.
  2. Set out the next steps in order, with who does each one and by when.
  3. Explain what Tom should check in his account's settings and terms, and what deleting the chat can and cannot achieve.
  4. Recommend two or three changes that would stop this happening again without banning AI.

Aim for 400 to 700 words. Every module ends in a task like this one.

How it's marked

Each part of the grid is judged Not shown, Partly, Mostly or Fully. Fully earns all of that part's marks, Mostly two thirds, Partly one third, and Not shown none. The parts add up to 100. Answers are marked by AI against this grid, and a CIMA-qualified (CGMA) moderator checks a sample.

Identifies and separates the risks (30 marks)

Fully
Separates all three: personal data of contacts and about 30 sole traders, including notes about an owner's private life, disclosed to a provider with no company contract; supplier spend and terms that are commercially confidential; and bank details that could be used for payment redirection fraud. Notes that the company, not Tom, is responsible as controller.
Mostly
Separates at least two of the three risks. Recognises personal data (contacts or sole traders) but misses either the sensitive notes or the fraud risk from bank details.
Partly
Risk is described in general terms ('data was leaked') without separating data protection, commercial and fraud risk. The sole traders and the notes column are not mentioned.

Sets clear, ordered next steps (30 marks)

Fully
Gives an ordered sequence with owners and times: record facts and settings with screenshots today; contain (delete chat, files and the emailed copy, switch off training); tell the data protection lead today, who decides on reporting within 72 hours of Thursday's awareness and records the breach either way; extra checks on supplier bank detail changes; consider whether supplier contracts require notice.
Mostly
Gives a reasonable sequence (record, contain, inform) with some owners, but misses the 72-hour clock from when the company became aware, or does not hand the reporting decision to the data protection lead.
Partly
Steps are missing, out of order, or limited to 'delete the chat'. No owners or timing.

Knows what to check in the account and what deletion does (20 marks)

Fully
Lists specific checks: the training or model improvement setting and its state on Tuesday, chat history and retention, feedback given, where data is stored, and the provider's deletion period. Explains that deletion removes the chat from history but the provider may keep copies for a period, and data already used for training cannot be pulled back.
Mostly
Says to check the training setting and delete the chat, but does not explain retention after deletion or that past training cannot be undone.
Partly
Assumes deletion removes the data, or assumes a paid plan is a business plan.

Prevents a repeat without banning AI (20 marks)

Fully
Gives two or three specific measures tied to the cause: a one-page rule on which data goes into which tool; the business assistant made easy to reach on phones; a minimised supplier extract (no bank or contact columns) for this kind of analysis; a no-blame route for reporting mistakes quickly.
Mostly
Gives one or two sensible measures, such as a policy, but does not tie them to the cause (no approved route was easier than a personal app).
Partly
Recommends a ban, discipline only, or nothing specific.

An answer

Illustrative: an invented answer, not a real learner's work.

To: Aisha Bello, Finance Director
From: Management accountant
Subject: Supplier list in a personal AI account: risks and what we do today

1. The risks
Data protection. The list holds personal data: contact names, emails and mobile numbers, and about 30 suppliers are sole traders, so their spend and bank details are personal data too. The notes column is worse: a comment about an owner's divorce is private information about a named person. It has gone to a provider we have no contract with. We are the controller, so this is the company's breach to handle, not Tom's.

Commercial confidentiality. Annual spend and payment terms for 140 suppliers tell anyone where our leverage is. Some supplier contracts may also require us to keep their terms confidential.

Fraud. Sort codes and account numbers are what a fraudster needs to impersonate a supplier or redirect a payment. Even if nothing has leaked, we should treat bank detail changes with extra care for the next few months.

2. Next steps
- Today, Tom: screenshot the chat, the account's data settings and the plan name, then delete the chat and the emailed ranking. Switch off any training or model improvement setting.
- Today, you: tell our data protection lead, with Tom's screenshots and a short timeline.
- This week, data protection lead: decide whether this must be reported to the ICO and record the breach either way.
- From today, AP team: call back every supplier asking to change bank details, on the number we already hold.
- This week, purchasing: check whether any supplier contract needs us to tell them.

3. What Tom should check
- The training or model improvement setting, and what it was on Tuesday, not just now.
- Chat history and how long the provider keeps chats, including after deletion.
- Whether he gave feedback on any reply, as some providers keep those conversations longer.
- Where the provider stores data, from its privacy terms.
Deleting the chat removes it from his history, but the provider may keep a copy for a set period. If the chat was used for training, that cannot be undone.

4. Stopping it happening again
- A one-page rule on which data may go into which tool, with examples from our own files.
- Put the company's AI assistant on staff phones. Tom used his own app because ours was harder to reach.
- A cut-down supplier extract for this kind of analysis, with no bank or contact columns.
- A no-blame way to report mistakes quickly. Tom told us within two days; we want that every time, and faster.

Its feedback and mark

90

out of 100. A pass: 80 or more.

Identifies and separates the risks Fully · 30 of 30

You separate all three risks and tie each to the list: contacts and the 30 sole traders, the divorce note as private information, commercial terms, and bank details as a fraud route. You also say the company is the controller.

Sets clear, ordered next steps Mostly · 20 of 30

A clear order with owners: record, contain, inform, then the bank-change and contract checks. But you give the reporting decision 'this week'. The 72-hour window runs from Thursday morning, when the company became aware, so the data protection lead needs to decide by Sunday morning.

Knows what to check in the account and what deletion does Fully · 20 of 20

Specific checks, including the setting as it was on Tuesday and any feedback given. You explain plainly what deletion does and does not do, and that past training cannot be undone.

Prevents a repeat without banning AI Fully · 20 of 20

Three measures tied to why it happened: a data rule with real examples, the company assistant on phones, and a minimised extract. The no-blame reporting route is a good fourth.

What worked

  • Written as a note the FD can act on today, with an owner against every step.
  • Risks are specific to this list, not general warnings.
  • The prevention steps deal with the cause: the approved tool was harder to reach than a personal app.

To lift your mark

  • Put the 72-hour clock in the next steps: the data protection lead decides on reporting inside the 72-hour window from Thursday morning, not 'this week'.
  • Say who contacts the provider about deletion if the settings show the chat may have been kept.

Under 80, you use the feedback and submit again. Your best mark counts. After a pass you also see a model answer.

Reflect, then copy it to your CPD record

After each task you note what you learned and how you will use it. One button lays it out as an entry for your CPD record (CIMA's, another body's or your employer's), in your words:

Illustrative reflection: “A personal paid plan is not a business plan: the training setting, retention and contract terms are different, and the company is still the controller. Deleting a chat does not undo what the provider has already kept or used.”

Development objective:
To be able to explain how consumer and business AI accounts differ on training, retention, storage and access, check those points in a plan's terms, and lead the response when company data has been put into a personal account.

Activity:
Consumer v business accounts: retention, training and where data is stored (Data security and governance for AI in finance)

Provider:
Eighty Marks (CPD)

Date completed:
1 November 2026

Evidence:
Written task marked 90 out of 100 (pass mark 80), with feedback

What I learned:
A personal paid plan is not a business plan: the training setting, retention and contract terms are different, and the company is still the controller. Deleting a chat does not undo what the provider has already kept or used.

How I will apply it:
I will draft our one-page data rule for the AP and purchasing teams, and ask IT to put the company assistant on work phones this month.

The certificate

Pass every module of a course at 80 or more and you get a certificate with a link you can share. It names each module's marked task and your mark.

Eighty Marks · CPD · Sample

Sample Learner

completed Data security and governance for AI in finance, passing every module's marked written task at 80 or more out of 100.

  • What can go in: public, internal, confidential and personal data86
  • Consumer v business accounts: retention, training and where data is stored90
  • UK GDPR, confidentiality and client data91
  • Writing a team AI policy and keeping an audit trail88

Illustrative: an invented learner and marks.

What an employer sees

When an employer pays for places, their managers see marks, passes, the traps missed in labs, and certificates. Never anyone's written answers.

LearnerModules passedBest markLab traps missedCertificate
Learner A (AP)4 of 488NoneYes
Learner B (AR)3 of 481A deduction with no reasonNot yet
Learner C (Purchasing)1 of 472Duplicate remittance; currency bank chargeNot yet

Illustrative: an invented team. Managers can also download it as a CSV.

Start with this course

Data security and governance for AI in finance: 4 modules, each with a marked task like this one.

See the course · Buying for a team