Module 4 – Using AI safely and smartly
Moving forward with confidence
Total time: about 40–55 minutes · Roughly 15 minutes reading, 25–40 minutes practice
You learn to use AI safely and sensibly: what to share and what not to, how to check answers on important topics, and what limits are realistic. This is the fourth of five modules: here you lay the groundwork for responsible use; in module 5 you apply that to medical questions in ChatGPT. You bring a lot together: not only “using it handily”, but safely, with the right expectations and control over what you get back.
Settings and “when in doubt, do not paste”
In ChatGPT, under Data controls / privacy settings, you can influence what happens with your conversations. Temporary chat can be useful when you want to try something without it behaving like a normal saved thread, exact names and menu locations change; check OpenAI’s official help shortly before you teach or take the course.
Free vs paid: the core of this module is behaviour (what you share, what you verify). Extras such as business workspaces or higher limits are optional paid and only matter if your organisation offers them.
Rule of thumb: if you are unsure whether something is sensitive, do not paste. Prefer a fictional example or only the question without identifiable details.
Lesson 4.0 – The 3 ground rules (if you only remember this)
Goal: three simple rules that always work, even when you are short on time.
Using ChatGPT does not have to be scary. You only need a few simple habits. These three rules almost always keep you on safe ground:
- Rule 1: Do not share secrets (passwords, codes, private documents, full personal details).
- Rule 2: For important topics: always double-check (health, money, legal, news).
- Rule 3: You decide: you use AI as a helper, not as the final boss.
Health: ChatGPT can help with explanations, understanding letters, or thinking of questions for the doctor (as in Module 1). It is still not a doctor, for real symptoms or doubt your GP or other care is still the place to go.
Quick read: what is sensible?
Usually fine: you ask for help with language or tone, for example: “Make this sentence a bit friendlier.” As long as you do not paste secret things in that message, no PIN, no full address, no password, that is generally okay.
Rather not: your full card number + date of birth + “what should I fill in?” in one message.
In between (smart): replace names and numbers with [neighbour], [invoice number], then you can still get help without exposing everything.
Quick check
1. What is the best “default mindset” with AI?
2. Which task is riskiest to follow blindly?
Lesson 4.1 – Privacy: what to share, what not?
Goal: learn to work “safe enough”, without paranoia.
The main skill in this lesson is simple: you learn the difference between useful context and too personal information.
What can you usually share safely?
- A general situation without names (e.g. “I want an email to a colleague”)
- Text you would send publicly anyway
- Questions about explanation, summarising, spelling, tone
What should you rather not share?
- Passwords, PINs, recovery codes, API keys
- Copies of passport/ID, national insurance numbers, full address details
- Full medical records or documents with name + date of birth
- Trade secrets or confidential contracts (especially with amounts/client names)
Handy trick: “Make it anonymous”
If you do need help with a text, you can almost always “clean” it. Replace names with [colleague], [school], [company], and leave out sensitive numbers.
Questions for lesson 4.1
3. What is the best approach if you want a rewrite but the text contains names?
4. Which information does not belong in your message to ChatGPT?
Mini exercise (fill-in)
List 3 things you should never paste into ChatGPT (for example: password).
Lesson 4.2 – Reliability: how do you check an answer?
Goal: learn a simple “fact-check routine” when something matters.
ChatGPT can sometimes make up something that sounds very convincing. In plain language that is: “an answer that sounds logical but is not true”. That is not malice; it is simply how a language model works.
The 5-step check (easy to remember)
- Step 1: Ask about certainty: “How sure are you? What are the doubts?”
- Step 2: Ask for sources: “Which reliable sources can I use for this?”
- Step 3: Check 1–2 core facts yourself (e.g. official website, government, doctor, bank, manual).
- Step 4: Let ChatGPT critique itself: “Name 3 possible mistakes in your own answer.”
- Step 5: Make it practical: “Make a checklist so I can do this safely.”
Questions for lesson 4.2
5. What is a sign you should double-check extra?
6. Which question often makes an answer more reliable right away?
Mini exercise (fill-in)
In your own words, name 2 steps you take when something is important and you want to check an AI answer.
Lesson 4.3 – Boundaries: when to be extra careful
Goal: recognise situations where you use AI only as extra help.
Some topics are “high risk”. Then ChatGPT is fine for explanation or preparation, but not as the final word. That applies especially to:
- Medical symptoms and medication
- Legal steps (contracts, fines, appeals, divorce)
- Money and investing (taxes, loans, large purchases)
- News, war, politics (can change, can be biased)
How can you still use AI smartly here?
- Let it help you phrase questions for your GP/lawyer/bank
- Let it explain what words mean in simple language
- Let it summarise what you already have (e.g. a letter you received, without private details)
Questions for lesson 4.3
7. What is the smartest way to use ChatGPT for health topics?
8. Which sentence helps put AI in the right role?
Lesson 4.4 – Other language models (short and simple)
Goal: see that “AI” is not one brand, and what usually differs.
ChatGPT is one well-known AI chat. But there are several “language models” (LLMs). The idea is the same: you type, you get text back. These are well-known examples:
For beginners what matters most: pick one and learn the basics. If you get curious later, you can paste the same prompt into two tools and compare. That is also a smart check: if two models disagree, you verify more carefully.
Questions for lesson 4.4
9. What is a smart reason to compare two AI tools sometimes?
10. What matters most for a beginner?
Lesson 4.5 – Your fixed prompt template (you can always copy this)
Goal: a “recipe” that almost always works, without fuss.
Many beginners get stuck because they think they must ask “the perfect question”. You do not have to. Just use this template and fill it in.
When you are ready, send that prompt. If the answer is not perfect, do what you already learned: steer it. For example: “shorter”, “simpler”, “friendlier”, “with an example”.
Questions for lesson 4.5
11. Why does a template often work well?
12. What do you do if the answer does not fit well?
Final tasks – finish module 4
These are not exams. They are “level-ups”. If you do these five, you round off the safety and trust foundation of this module. Then you move on to module 5. Take your time.
Click here to open ChatGPT in a new tab, this course page stays open.
Task 1 – The perfect WhatsApp (2 versions)
Type in ChatGPT (copy and fill in):
Task 2 – Make a tricky email safe (no names)
Type:
Task 3 – “Check yourself” prompt
Type:
Task 4 – The “calm explanation” prompt
Type (and pick a topic you find difficult):
Task 5 – Your personal AI rules (mini checklist)
Type:
Last practice check (fill-in)
Write one sentence you will use from now on to “set limits” for ChatGPT (e.g. explanation yes, medical advice no).
Write your own mini-template (short) for a situation you often have. (Situation + goal + format + tone.)
Closing – Safety foundation done
You can now:
- use AI without oversharing
- judge answers and double-check important things
- put AI in the right role (explanation/ideas/text, not “final advice”)
- build good prompts quickly with a template
You do not have to be “technical”, only know how to use it smartly and safely. And if you ever doubt: share less, check more, and steer. Next step: in module 5 you practise medical-style questions in ChatGPT; evaluation comes after module 5.