Lesson 3 of 4 · about 5 min
Guardrails for financial analysts
When you finish, you can protect confidential and market-sensitive data and keep AI inside your firm's and regulator's rules.
Stuck? Ask the AI tutorWhat never goes into an unapproved tool
Never paste client names, account details, portfolio positions, non-public results, deal information, payroll data or anything marked confidential into a consumer AI tool unless your employer has approved that tool for that data. Market-sensitive information can create serious problems if it leaks, even by accident.
Many employers have rules that tie to securities laws, data-protection laws and confidentiality agreements. Learn yours. If you do not know whether a tool is approved, ask your manager or compliance team before you use it, not after.
Accuracy and the duty to check
AI can misstate numbers, mix up periods, confuse similar company names and invent figures that look plausible. A summary that says revenue rose when it fell is worse than no summary. Check every number that matters back to the source document or system.
Keep an audit trail. Note which tool was used, what it did and how you verified it. If a regulator, auditor or manager asks how a number was produced, you should be able to show your steps.
Not advice, and bias in the output
In regulated settings, investment, tax and accounting advice come from qualified professionals and licensed firms, following the rules of their regulator. An AI tool cannot hold a license or take responsibility. Do not present AI output as a recommendation. This course gives no financial advice.
Bias can creep in. Training data may overstate certain markets, ignore smaller companies or reflect old assumptions. Models used for credit, hiring or pricing can treat groups unfairly. If your work touches decisions about people, ask about fairness testing and human review.
Disclosure and saying no
Follow your firm's rules about disclosing AI use to managers, auditors and clients. If a report relies on AI assistance, say what was done and who reviewed it. Do not pass off a draft as your own verified analysis.
Say no when asked to use unapproved tools on confidential data, to skip checks because of time pressure or to present an AI-generated number as fact. A calm line works: “I can have a draft by noon, but I need time to verify the figures against the ledger.”
Practice now
- Write a personal rule list: three kinds of data you will never paste into an unapproved tool.
- Make a verification log template with columns for tool, task, source checked and reviewer.
Use any AI assistant you have. Do not paste private, client or patient information into it.
Check yourself
1. You have unreleased quarterly results. Where may you put them?
2. An AI summary says costs fell. The source table shows they rose. What does this show?
3. Who is responsible for a recommendation to a client?
Ask the AI tutor
Ask anything about this lesson, in any language. The tutor is an AI: it can be wrong, and it never needs your personal details.
AI-drafted. Checked by an automated validator and a separate AI reviewer; a person has not read every lesson. Translations and tutor answers are AI-made. Found a mistake? Tell us. AI can be wrong. General information only, not professional advice. Not affiliated with any AI company named. Current University is not accredited.