IntegrAuth Academy
Identity, security and practical AI — free, from zero. Byte-sized lessons, real diagrams, hands-on labs, zero fluff. Two programs, each with its own exam and certificate.
English is hard? Your browser can translate this page into Tamil, Telugu, Kannada, Malayalam or Hindi. Tech words stay in English so you learn the real names.
How?
- Chrome on phone: ⋮ menu → Translate
- Chrome on computer: right-click → Translate to …
- Edge: the translate icon in the address bar
Cheat sheet & pop quiz
You finished the track. Here it is on one page, plus 5 quick questions. Your goal: answer each one before you reveal it.
If you remember nothing else
| Lesson | The idea to keep |
|---|---|
| Prompt injection | A better prompt can't fully fix prompt injection. Watch for the lethal trifecta: private data, untrusted content and a way to send data out. Keep approvals on, and read them. |
| Deepfakes and AI scams | You can't reliably spot a good fake. Urgent + secret + odd payment means hang up. Then check on a channel you choose. Get AI apps from the real maker. A missing label proves almost nothing. |
| Bias and fairness | Bias comes in through the data, the labels and how the tool is used. Deleting a field leaves proxies behind. Run a swap test on your data. |
| Who owns AI output? | Ask three questions: who made it, does it copy anyone, and what do the terms say. In the US, copyright needs a human author. Other countries differ. This is not legal advice. |
| AI rules at work | Unapproved AI shows what people need. Make a safe path. Write a short policy that says yes to safe uses. Keep a list of AI in use, each with an owner. The EU AI Act sorts AI by what it is used for. |
| Staying sharp with AI | AI that is usually right makes people stop checking. This is automation bias. Put the source beside the suggestion. Spot-check. Use the smallest model that does the job. |
| AI with children and families | Make the rules together. Homework: hints, not answers. Check each app's age limit and settings. A bot is not a friend. A fake nude: don't share, tell, report. |
Pop quiz: 5 questions
Q1 · Sam connects his assistant to his work inbox and to a web-browsing tool. Then he asks it to research a supplier. He says, "It only reads things, so it is safe." Is he right? What is the smallest change that helps?
No. His inbox holds private data. It also holds untrusted content, because anyone can email him. The browsing tool is a way out. Every page it opens can carry data out. That is the whole trifecta in one chat. Do the research in a separate chat with no inbox connected. Then a fooled assistant has nothing to steal. See prompt injection.
Q2 · Maya gets a video call from a "police officer". The caller wears a uniform and sounds real. They say Maya is under "digital arrest". She must move her savings to a "safe account" tonight and tell no one. What should she do?
End the call. Police and courts do not arrest anyone by video call. Then tell a family member, and call 1930 or visit the police station herself. Never use a number or link from the call. The call is urgent and secret, and it asks for an odd payment. A real-looking face or uniform proves nothing. See deepfakes and AI scams.
Q3 · A vendor says, "Our AI screener can't be unfair about age. It never sees birth dates." What would you test before you use it?
Test for proxies. A graduation year can stand in for age. Run a swap test. Use the same résumé twice, and change only the graduation year. Run it several times. Then compare the results for each age group on your own applicants. Keep a recruiter who can say no on every rejection. See bias and fairness.
Q4 · Priya's team wants a new logo. The chatbot's terms say "you own the output." A colleague wants to make one from a prompt and register it. What is missing?
The terms can only pass on rights that exist. In the US, a logo made from a prompt alone likely has no human author. Then there may be no copyright to own. The terms also don't tell you if it copies someone else's work. For an important logo, add real human design work. Check for look-alikes, and ask a lawyer. Rules differ by country. See who owns AI output?
Q5 · An AI checks invoices. After a year, reviewers approve almost everything it marks "no issues". Nobody has disagreed with it in months. Does this prove the tool is excellent?
Not by itself. When a tool is usually right, people stop checking. This is automation bias. No disagreement can mean nobody is really looking. Problems the tool never flags slip through. Show the invoice beside the AI's result. Spot-check "no issues" items. Add checks the tool was not built for, like duplicates. The person who approves still owns the result. See staying sharp with AI.
Go deeper: a checklist before you rely on AI
- What is this assistant connected to? What can it do without asking? See prompt injection.
- If this request is fake, which second channel would show it? See deepfakes and AI scams.
- Does the output decide something about a person? Did anyone test it for bias on our data? See bias and fairness.
- Will we own or publish this? Does it copy anyone? See who owns AI output?
- Is the tool approved and on our list, with an owner? See AI rules at work.
- Could I still catch it when it is wrong? Is this the smallest model that will do? See staying sharp with AI.
- Does my child know what to do with a fake picture? See AI with children and families.
That's Safe and responsible AI. You can limit what a fooled assistant can reach. You can check a request on a channel you choose. You can test a tool for bias. You can tell who owns AI output. You can write an AI policy people follow. You can keep your own judgment in charge. And you can set AI rules with your family. Next up — Coding with AI: using coding assistants without losing control of your code. Start at Start here: coding with AI.
New to files, the terminal and git? Read Tech basics: terminal, JSON and keys first. That track assumes it.
List every app connected to the assistant you use. Switch off one you don't need. Agree on a code word with your family. Write down one work decision where AI is used, who owns it and who checks it.
Want more? Browse our free security tools on the tools shelf.