Day 7: What to Keep Private

The Concept

Using AI well includes knowing what not to share with it.

Most everyday AI tasks do not require sensitive information. You can ask for help writing, summarising, planning, explaining, brainstorming, or organising without pasting private details into the chat. The mistake many people make is not malicious. It is casual. They copy too much context because it is faster than pausing to remove what the tool does not need.

Today is about building that pause.

Today's goal: learn how to give AI useful context without exposing private, personal, or confidential information.

This does not mean you should be afraid of AI. It means you should treat AI like any other tool that handles information. Share what is useful. Remove what is unnecessary. Follow your organisation's rules when work information is involved.

The privacy principle

The safest everyday rule is simple:

Share the shape of the situation, not the sensitive details.

AI usually does not need real names, exact figures, account numbers, private addresses, client identifiers, unreleased strategy, or confidential documents to help you think or write. It needs the role, relationship, goal, tone, constraints, and type of outcome.

For example, AI does not need:

• "Sarah Johnson from Acme Ltd has not paid invoice 48291 for $14,275."

It can work with:

• "A client has not paid a mid-sized invoice that is now overdue. I want to send a polite but firm follow-up."

The second version gives enough context to produce a useful draft while reducing unnecessary exposure.

What to keep out

Some information should almost never be pasted into a consumer AI chat unless your organisation has explicitly approved the tool and use case.

1. Personal identifiers

Keep out information that can identify a person in a sensitive way:

• full names combined with other private details,
• home addresses,
• dates of birth,
• government ID numbers,
• passport or visa details,
• medical record numbers,
• personal phone numbers,
• private email addresses,
• employee IDs when unnecessary.

Plain English: If the detail could help someone identify, impersonate, or profile a person, remove it unless it is truly required and approved.

Most prompts work perfectly well with placeholders like "a colleague," "a client," "my manager," "a family member," or "a customer."

2. Financial and access information

Never paste passwords, PINs, API keys, recovery codes, bank account details, card numbers, or login credentials into AI.

There is no everyday writing, planning, or learning task that requires those details. If a task seems to require them, redesign the task. Ask for a template, explanation, or checklist instead of sharing the secret itself.

Use this rule: AI can help you write about a password process. It should not receive the password.

3. Confidential work information

Professional context needs care.

Avoid pasting:

• client names when not needed,
• private customer data,
• unreleased product details,
• internal strategy,
• legally sensitive material,
• confidential HR information,
• proprietary code or documents,
• anything covered by an NDA or workplace policy.

If your organisation has an AI-use policy, follow it. If you are unsure, ask. The cost of checking is small compared with the cost of mishandling confidential information.

Workplace habit: Before using AI for work, know which tools are approved and what kinds of data are allowed.

What is usually fine to share

The good news is that most useful AI prompts do not require sensitive data.

It is usually fine to share:

• general descriptions of a task,
• public information,
• your own non-sensitive writing,
• a summary you created without private details,
• goals and constraints,
• tone preferences,
• anonymised scenarios,
• rough examples with placeholders.

AI is often just as useful when the details are generalised.

Instead of sharing:

• "My manager Priya said my Q3 performance rating may drop because of the Orion account."

Use:

• "My manager gave me feedback that my performance may be rated lower because of one project. Help me prepare for a constructive follow-up conversation."

The second version protects names and business context while preserving the useful shape of the problem.

How much detail is enough?

The goal is not to make your prompt vague. The goal is to make it safe.

Many people remove too much information once they start thinking about privacy. They replace every useful detail with a generic phrase, and then the AI response becomes bland. A better approach is to keep the context that affects the quality of the answer while removing the details that identify a person, account, company, or confidential situation.

For example, "I need help writing to someone" is too vague. "I need help writing to a client who is frustrated about a delayed project, and I want the tone to be calm, accountable, and practical" is useful without exposing private information.

Good privacy practice: Keep the relationship, goal, tone, and constraint. Remove the identity, secret, exact figure, and confidential fact.

That is the balance you are practising today.

The redaction workflow

Use this quick workflow before pasting context into AI.

Step 1: Write the full version privately

Start by writing what you actually mean in your own notes. This helps you think clearly.

Do not paste it yet.

Step 2: Replace sensitive details

Swap specifics for general descriptions:

• real names -> "a colleague," "a client," "my manager"
• exact amounts -> "a small amount," "a mid-sized invoice," "a significant budget"
• company names -> "my employer," "a vendor," "a client organisation"
• dates -> "next week," "last quarter," "recently," if exact dates are not needed
• private documents -> a short summary of the relevant part

Step 3: Keep the useful context

Do not remove so much that the prompt becomes vague. Keep the information AI needs:

• relationship,
• goal,
• audience,
• tone,
• constraints,
• desired output,
• what you have already tried.

The balance: Remove identifying details, not meaning.

Use this today

Take one task you might want AI to help with this week.

Write the context twice:

1. Full private version: write it for yourself, with all the real details.
2. AI-safe version: replace names, exact figures, IDs, confidential facts, and private details with general descriptions.

Then use only the AI-safe version in your prompt.

Notice something important: the safer version will usually still work. That is the lesson. Good prompting is not about giving AI every detail. It is about giving the right details.

Remember this

If you remember nothing else from Day 7, remember these three ideas:

• AI needs useful context, not unnecessary private details.
• Use placeholders and general descriptions before pasting sensitive information.
• For work tasks, follow your organisation's approved tools and policies.

Privacy is not a barrier to using AI. It is part of using AI well.