BEGINNER GUIDE · 14 MIN READ

AI safety for beginners: 7 simple rules

You do not need to be afraid of AI. Protect private information, check important answers and keep a human involved when the consequences matter.

7 practical rulesNo technical knowledge neededQuick checklist included
The one rule to remember

If you would not post it publicly or email it to a stranger, stop before putting it into an AI tool.

01

What safe AI use really means

Safe use does not mean avoiding AI. It means matching your level of care to the possible harm. Asking for dinner ideas is low risk. Uploading a medical letter, approving a contract or sending money is not.

Green: usually low risk

Brainstorming, rewriting your own text and learning general ideas.

Orange: pause and check

Uploading files, using work data or relying on facts that may change.

Red: get human help

Emergencies and decisions about health, law, money, safety or someone’s rights.

02

The seven simple rules

1. Keep secrets secret

Never enter passwords, PINs, recovery codes or security answers.

2. Limit personal details

Share only what is necessary, and remove identifying information where possible.

3. Inspect every upload

Check documents and images for hidden or sensitive information.

4. Verify important facts

Open the sources yourself and prefer official information.

5. Keep humans in charge

Use qualified help for decisions with serious consequences.

6. Review before sharing

Check text, images and claims before you send or publish them.

7. Use the right tool

Check its privacy settings, age limits and intended use first.

03

Protect private information

An AI chat is an online service, not a private notebook. Give the tool only the information it needs. Replace real details with neutral labels such as “Client A”, “my city” or “£X”.

Do not paste
  • Passwords, PINs or one-time codes
  • Passport, tax or national ID numbers
  • Bank or card details
  • Private health, HR or customer records
  • Another person’s information without permission
Safer version

Instead of: “Reply to Maria Santos at this address about account 48392.”

Try: “Draft a polite reply to a customer about a delayed order. Use placeholders for the name and order number.”

Passwords are never needed for a normal prompt.

If a chatbot asks for one, stop. Open the service’s official website or app yourself.

04

Check files, photos and screenshots before uploading

A document may contain more than the paragraph you want help with. Look at every page and remove details the AI does not need.

Look for

  • Names, addresses and contact details
  • Signatures, faces and location clues
  • Account, invoice and reference numbers
  • Tracked changes, comments and hidden tabs
  • Company, school or customer information

A safer workflow

  1. Make a copy of the file.
  2. Remove or cover private details.
  3. Upload only the pages you need.
  4. Delete the upload or chat afterwards if appropriate.

Do not upload confidential work, school, customer or patient material unless your organisation has approved that specific tool and use.

05

Check important answers instead of trusting the tone

AI can produce an answer that sounds clear and confident even when a date, name, quotation or source is wrong. This is often called a hallucination. Confidence is a writing style, not evidence.

Ask

“Which claims in this answer should I verify, and where can I check them?”

Open

Visit the cited page yourself. A source title or link can also be invented or misread.

Compare

For important claims, use an official source or two reliable independent sources.

Check the date

Prices, rules, software features and public information can change quickly.

A useful verification prompt

List the factual claims in your answer. For each one, give an official source, its publication date and anything you are uncertain about.

06

Know when an AI answer is not enough

AI can explain unfamiliar words and help you prepare questions. It should not make a final decision that could seriously affect your health, legal rights, money, employment or personal safety.

Health

Use AI to understand general terms. Use a qualified professional for diagnosis, medication and treatment. In an emergency, contact local emergency services.

Law and rights

Use AI to organise questions. Check current law and deadlines with an official source or qualified adviser in your location.

Money

Use AI to learn concepts. Verify rates, fees, tax rules and investment risks before acting.

Jobs and people

Do not let an unreviewed AI score decide who is hired, disciplined, graded or given access to a service.

07

Recognise AI-assisted scams and deepfakes

A voice, image or video is no longer proof by itself. Scammers can copy a familiar voice, imitate a manager or create a believable message to make an urgent request feel real.

Red flags
  • Unexpected urgency, secrecy or threats
  • A request for money, gift cards, crypto or login codes
  • A new phone number or unusual payment method
  • A link or attachment you were not expecting
  • Pressure not to call anyone else
Stop

Do not pay, click or share information while you feel rushed.

Verify separately

Call the person or organisation using a number you already know.

Use a family check

Agree on a private question or word for unexpected emergency requests.

Report

Tell the platform, bank, employer or relevant local authority when appropriate.

08

Use extra care at work, school and with children

Work and school

  • Follow your organisation’s AI policy.
  • Do not upload protected or confidential material.
  • Check facts, citations and calculations.
  • Be honest when disclosure of AI use is required.
  • Keep a person responsible for the final result.

Children and families

  • Use age-appropriate tools and account settings.
  • Teach children not to share names, photos or locations.
  • Explain that AI can sound human without being a person.
  • Review outputs together and discuss mistakes or bias.
  • Keep trusted adults involved in worrying conversations.
09

Your before-you-send checklist

Pause for ten seconds and ask:
  1. Does this contain a password, code or financial detail?
  2. Does it identify me or another person unnecessarily?
  3. Do I have permission to share this file or information?
  4. Could a wrong answer cause real harm?
  5. Will I check important facts in the original source?
  6. Would I be comfortable if this information became public?

If any answer worries you, remove details, use a safer example, choose an approved tool or ask a trusted person for help.

10

Common questions

Can an AI tool keep my chats private?

Privacy rules differ between tools, plans and account settings. Do not assume a chat is confidential. Check the tool’s current privacy controls and avoid entering information that would cause harm if it were exposed.

Is it safe to upload a document?

It can be, if you have permission and first remove sensitive details. Check names, addresses, signatures, account numbers, hidden comments and confidential work information before uploading.

Can I ask AI a medical question?

You can use it to learn general terms or prepare questions for a professional. Do not use an AI answer as a diagnosis or as the only basis for treatment, medication or an emergency decision.

How do I know whether an AI answer is true?

Ask for sources, then open those sources yourself. Check important claims against an official website or more than one reliable source. A confident tone is not proof.

What should I do if an AI-generated voice asks for money?

Stop and verify through a different channel. Call the person on a number you already know, ask a question only they would know, and never send money because a call creates urgency.

Can children use AI safely?

They need age-appropriate tools, clear family or school rules and adult guidance. Teach them not to share personal details, photos, school information or another person’s private information.

What if I already shared sensitive information?

Delete the chat if the service allows it, review the account’s data controls, and change any exposed password immediately. If bank, identity, medical or work data was involved, contact the relevant organisation and follow its security advice.

KEEP LEARNING

Use AI with confidence, not blind trust

Start with low-risk tasks, share less information than you think you need, and keep the final decision in human hands.