AI safety, ethics, privacy and your career
AI is changing how people work, learn and get information. Using it responsibly protects you, the people you work with, and your future career. This lesson covers the risks, the ethics, and how to stay valuable in an AI world.
How AI chatbots work (in one paragraph)
Large language models (LLMs) learn patterns from huge amounts of text and predict likely next words. That's why they're fluent and helpful, and also why they can be confidently wrong: they generate plausible text, which isn't the same as checked facts. Newer tools can search the web, read files or run code, but you're still responsible for the result.
Main risks and what to do
| Risk | What it looks like | Protect yourself |
|---|---|---|
| Wrong information | Invented facts, laws, fees, references | Verify with official or primary sources |
| Privacy leaks | Pasting ID numbers, client data, passwords | Don't share personal or confidential data; check privacy settings |
| Bias | Unfair assumptions about gender, ethnicity, age | Question outputs; review decisions affecting people |
| Deepfakes | Fake voices and videos of real people | Verify unusual requests by calling back on a known number |
| AI-powered scams | Perfectly written phishing, cloned voices of relatives | Use a family "safe word"; never send money on a voice note alone |
| Over-reliance | Can't do the work without AI | Keep practising core skills |
| Copyright | Copying protected text, images, code | Use your own work and properly licensed material |
Deepfakes and voice cloning
Scammers can clone a voice from a short clip and call pretending to be a child, parent or boss asking for urgent M-Pesa. Defences:
- Hang up and call back on the number you already have.
- Agree a family code word for emergencies.
- Be suspicious of urgency and secrecy.
- Remember video and audio can be faked; check with other sources.
Ethics: using AI fairly
- Be honest about AI use in school and work where it matters.
- Don't harm or deceive: no fake reviews, impersonation, harassment or misinformation.
- Respect privacy and consent, including other people's photos and data (Kenya Data Protection Act).
- Keep humans in charge of important decisions about people: hiring, loans, grades, medical care.
- Credit sources and respect creators.
Using AI at work
- Check your employer's AI policy before using tools with company data.
- Use approved/business versions for confidential work where available.
- You are accountable for anything you send, publish or deploy, even if AI drafted it.
- Keep records of important AI-assisted decisions.
Your career in an AI world
AI is best at tasks; people are still needed for judgement, relationships, context and responsibility. To stay valuable:
| Build | Why |
|---|---|
| Strong fundamentals (maths, writing, coding, networking) | You can check and improve AI output |
| Domain knowledge (agriculture, health, finance, education) | AI needs someone who understands the real problem |
| Communication and teamwork | Clients and managers trust people |
| AI skills: prompting, automation, evaluating outputs | Makes you much more productive |
| Ethics and good judgement | Increasingly valued as AI spreads |
Jobs growing with AI include AI-assisted content and design, data labelling and quality review, automation for small businesses, AI trainers and prompt/workflow specialists, and developers building AI into apps.
A personal AI checklist
- [ ] I verify important facts and test code.
- [ ] I don't paste private or confidential data.
- [ ] I follow my school's or employer's AI rules.
- [ ] I label or disclose AI use where it matters.
- [ ] I practise core skills without AI too.
Why AI safety and ethics concern everyone
AI is now part of hiring, lending, healthcare, education, news and government services. It can help people, but it can also spread misinformation, enable scams, leak private data, or treat people unfairly if used carelessly. Understanding the risks helps you protect yourself and your family, use AI responsibly at work, and take part in conversations about how AI should be governed.
How AI can be misused against you
| Threat | Example | Protection |
|---|---|---|
| Voice cloning scams | A "relative" calls in distress asking for urgent M-Pesa | Hang up and call back on a known number; agree on a family code word |
| Deepfake videos | A "celebrity" promotes an investment scheme | Check official accounts and credible news; be sceptical of "guaranteed returns" |
| AI-written phishing | Perfectly written emails pretending to be your bank or employer | Verify through official channels; don't click links in unexpected messages |
| Fake AI apps | "AI trading bot" apps that steal money or data | Install only from official stores; research the company |
| Impersonation chatbots | Fake customer care bots asking for PINs | Never share PINs, passwords or OTPs |
| Non-consensual images | Fake intimate or embarrassing images | Report to platforms and authorities; support victims; never share |
Bias and fairness
AI learns from data created by people, so it can repeat unfair patterns:
- A hiring tool trained on past hires might favour certain groups.
- Image generators may stereotype professions, regions or cultures.
- Language tools may work less well for Kiswahili, Sheng or other local languages.
Responsible use means: humans review important decisions, organisations test systems for unfair outcomes, and affected people can ask questions and appeal decisions.
Transparency: knowing when AI is involved
| Situation | Good practice |
|---|---|
| Customer service chatbots | Tell customers they're talking to a bot and offer a human option |
| AI-generated images in ads or news | Label them where they could mislead |
| AI-assisted school work | Follow and disclose according to school policy |
| AI in decisions about people (loans, jobs) | Explain how decisions are made and allow human review |
Privacy and data protection
- Don't share sensitive personal information (ID numbers, health records, passwords, financial details) with AI tools unless the tool is approved for it.
- Organisations using AI with personal data must comply with Kenya's Data Protection Act: lawful purpose, minimal data, security, and respecting people's rights.
- Check settings for chat history and training use; delete conversations you don't need.
Environmental and social costs
Training and running large AI models uses significant energy and water for data centres, and some AI systems rely on low-paid workers for data labelling. Using AI thoughtfully (for tasks where it adds real value) and supporting fair working conditions are part of responsible use.
AI at work: a responsible use policy (template)
- Approved tools: list which AI tools staff may use.
- Data rules: no confidential or personal data in unapproved tools; anonymise where possible.
- Accuracy: staff verify AI outputs and remain responsible for their work.
- Disclosure: when and how to disclose AI use to clients and colleagues.
- Prohibited uses: deepfakes, discrimination, plagiarism, automated decisions about people without review.
- Security: report suspicious AI-related scams; protect accounts with MFA.
- Training: staff learn safe and effective AI use.
- Review: update the policy as tools and laws change.
Preparing your career for an AI world
| Skills that grow in value | Why |
|---|---|
| Critical thinking and verification | AI outputs need checking |
| Communication and relationships | Clients and teams value human trust |
| Domain expertise | AI is more useful when guided by deep knowledge |
| Creativity and judgement | Choosing what matters, not just generating options |
| Digital and AI literacy | Using tools effectively and safely |
| Ethics and responsibility | Organisations need people who use AI properly |
Rather than competing with AI on speed, combine it with skills machines lack.
Talking to family about AI safety
- Explain voice cloning and deepfakes with simple examples.
- Agree on a family verification question for emergency money requests.
- Remind everyone: no PINs, passwords or OTPs to anyone, including "AI support".
- Encourage checking surprising news before forwarding it.
Practice
- Agree on a family code word and explain voice-cloning scams to someone older than you.
- Find an example of AI bias reported in the news and discuss how it could have been prevented.
- Review the privacy settings of the AI tools you use.
- Draft a one-page AI use policy for a small organisation using the template.
- List three skills you'll develop to work well alongside AI.
Think about it: A company uses an AI system to reject loan applications automatically, and applicants from certain areas are rejected much more often. What should the company do?Show answer
Investigate whether the system is unfairly penalising those areas (for example through biased training data or proxies for location), involve human review for rejections, test the model for fairness regularly, explain decisions to applicants and allow appeals, and comply with data protection and consumer protection requirements. Automated decisions affecting people's lives need accountability.
Check yourself
What does LLM stand for? (three words)
Show answer
large language model
A relative calls with a strange voice asking for urgent M-Pesa. What should you do first? (two words)
Show answer
call back
Should AI make final decisions about hiring people without human review? (yes or no)
Show answer
no
Who is responsible for work you publish that AI helped draft?
Show answer
me
What is a fake video or voice of a real person made with AI called?
Show answer
deepfake
What is an AI-generated fake video of a real person called?
Show answer
deepfake
Should customers be told when they're chatting with a bot? (yes or no)
Show answer
yes
What should families agree on to verify emergency money requests? (two words)
Show answer
code word