AI 101Beginner19 min read

Privacy, copyright and safe use: plain rules for using AI

A three-question check, a traffic-light rule for what to paste, a five-minute privacy audit, handling work and client material, copyright and ownership in plain terms, other people's faces, voices and data, honesty about AI use, and a one-page personal policy.

Privacy, copyright and safe use: plain rules for using AI

Privacy, Copyright and Safe Use: Plain Rules for Using AI

Most people learn the hard way. They paste a client contract into a chat to "just summarise it", share a link to a conversation that includes a colleague's personal details, or publish an AI image without wondering who owns it. None of these are reckless in the moment, and each can cause real trouble. This tutorial gives you plain rules for using AI sensibly: what never to paste, how to check a tool's privacy settings in five minutes, how to handle work and client material, what you can and cannot say about who owns what AI makes, how to treat other people's faces, voices and names, and when and how to say that AI was involved. It is practical, not legal advice, and it ends with a one-page personal policy you can keep.

Difficulty: ★☆☆☆☆ (no technical background needed)
Tools needed: Any AI tools you use, and access to their settings
Last updated: November 2026

Overview

Using an AI tool is not like using a calculator. When you type into a chat, you send text to a service run by a company. Depending on the tool, the plan and your settings, that text may be stored, reviewed by people in certain circumstances, used to improve the product, or shared with other services you have connected. When you ask the AI to make something, the result sits in a legal grey area that varies by country. And when you use AI on other people's faces, voices or data, you affect people who never agreed to it.

None of this is a reason to avoid AI. It is a reason to have a few simple habits, the way you have habits for email: do not forward a confidential file to the wrong person, do not click strange links, and think before you reply to all.

The habits come down to three questions, which you can ask in two seconds before almost any use.

1. What am I sharing? Is any of it private, confidential, personal or not mine to share?
2. Who can see or keep it? The tool, its staff in some cases, the people I share a link with, and other apps I have connected.
3. What could go wrong if it leaked, was wrong or was used the wrong way? If the answer is "something I could not undo", stop and rethink.

The tutorial has seven steps. First, those three questions in practice. Second, a simple traffic-light rule for what you can paste. Third, a five-minute privacy check of the tools you use. Fourth, how to handle work and client material. Fifth, copyright and ownership in plain terms. Sixth, other people: faces, voices, names and consent. Seventh, honesty about AI use, and a one-page personal policy to finish.

A note on what this is and is not. Laws on AI, privacy and copyright differ between countries and change often, and the year-in-review tutorial in this module shows how fast. Nothing here is legal advice. Where a rule matters to your job or business, check the official text or ask a qualified person. What follows is meant to keep you out of the common, avoidable trouble.

Who This Is Useful For

  • Anyone who uses AI at work, or for anything involving other people.
  • Freelancers and small business owners who handle client information.
  • Students and creators who publish AI-assisted work.
  • What You Will Learn

  • A three-question check you can run in seconds.
  • A traffic-light rule for what is safe to paste, and what never is.
  • How to audit the privacy settings of the tools you use in five minutes.
  • How to handle client and employer material, including anonymising it.
  • What can and cannot be said about copyright and ownership of AI output.
  • How to treat other people's likeness, voice and data, and how to avoid AI scams.
  • How to be honest about AI use, and a one-page policy you can keep.
  • What You Need

    1. The AI tools you use now, and a few minutes with their settings.
    2. If you use AI at work, your employer's policy, or a plan to ask about it.
    3. A notes app, for the one-page policy at the end.
    4. About thirty minutes.

    Information as of 14 November 2026. Settings screens, plan features, laws and court cases change often, and rules differ by country and by industry. This tutorial teaches principles and checks, and names no specific setting paths. Open your own tool's settings and the official sources for details, and treat anything here that matters to your work as a prompt to check.

    Step 1 — Three questions, in two seconds

    Before you paste, upload or ask, run the three questions.

    1. What am I sharing? Look at the text or file as if you had to explain it to its owner. Does it contain names, contact details, account numbers, health or financial information, passwords, internal plans or someone else's work? If you would feel uncomfortable reading it aloud to the person it concerns, it probably should not go in.

    2. Who can see or keep it? Think about the whole chain. The company that runs the tool may store your conversation, and some plans allow it to be used to improve the product unless you switch that off. If you connect the tool to your email, drive or calendar, it can read what those contain. If you share a link to a chat, anyone with the link may be able to open it. If you use a browser extension or a third-party app on top of an AI, that is another company in the chain.

    3. What could go wrong? Picture the worst plausible outcome: a leaked contract, a wrong claim about a real person, a published image that infringes someone's work, a scam call that sounds like your boss. Then ask how bad, and whether it could be undone. Mild and reversible: carry on. Serious or permanent: do not.


    Step 2 — A traffic-light rule for what you can paste

    Sort what you might share into three groups.

    LightWhat it coversRule
    Red: never pastePasswords, one-time codes, API keys and recovery phrases. Card, bank and government ID numbers. Other people's health, legal or financial records. Anything under a confidentiality agreement you have no clear permission to share. Unreleased company information such as source code, plans or results, unless your employer has approved the tool for itDo not put it in a general AI tool. Use an approved tool with proper terms, or do the task without AI
    Amber: think, then trimClient or customer details, colleagues' names, your own personal details, internal documents, draft contracts, other people's creative workRemove or replace the sensitive parts, check the tool's data settings and your employer's rules, and share only what the task needs
    Green: fine to pastePublic information, your own non-sensitive writing, generic questions, made-up examples, text you have rewritten with placeholdersNormal care. Still check the output before you rely on it

    Three further points.

    1. Credentials never belong in a chat. If an AI asks for a password or a code, stop. A genuine tool uses a proper sign-in, not a message in the conversation.
    2. Other people's data deserves the same care as yours. Names, phone numbers and details of customers, patients, students or colleagues are theirs. Being "just a few lines" does not change that.
    3. Images and files count. A screenshot or a photo can contain faces, names, addresses and documents in the background. Check what is visible before you upload.

    Step 3 — A five-minute privacy check of your tools

    Do this once for each tool you use regularly, and again after any big update. Open the settings and look for these.

    1. Training and improvement. Is there a setting that controls whether your conversations may be used to improve the product? Where is it, and what is it set to? You decide what you are comfortable with. If you want it off, switch it off now.

    2. History and temporary chats. Does the tool keep your history by default? Is there a temporary or incognito chat mode that does not save? Use it for sensitive questions where it exists, and read what "not saved" really means in the tool's own words, since some keep a copy for a short period for safety reasons.

    3. Deleting. How do you delete a chat, or all of them? How do you delete your account and data? Try deleting a test conversation to see that it works.

    4. Memory. If the tool remembers things about you, look at what it has stored, correct it and switch it off if you do not want it. Memory is convenient and is also a store of personal information.

    5. Connections. List every app the tool is connected to, such as email, calendar, drive and notes. For each, check what permissions it has and whether you still need it. Remove what you do not use.

    6. Sharing. If the tool lets you share a conversation by link, learn what that does. Treat a shared link as public to anyone who has it, and avoid sharing chats that contain anything sensitive.

    7. Plan differences. Personal and business plans often handle data differently. If you use a work account, find out which kind it is and what your employer's agreement says.

    8. Extensions and third parties. Review browser extensions, plug-ins and "AI wrappers" that sit on top of a model. Each is another company with access to what you type.

    Write down what you find and the date. Two minutes of notes now save you from guessing later.

    Step 4 — Work and client material

    If you use AI for work, the rules that matter most are your employer's and your clients'. Personal good judgement is not a substitute.

    Find out the policy. Many organisations have a rule on which AI tools are approved, what may be entered and what must be reviewed by a person. If you cannot find one, ask. "I want to use AI to speed up my drafts, which tools and what data are allowed?" is a perfectly normal question and shows good judgement.

    Use approved tools for confidential material. A company-approved tool under a business agreement is a different thing from a personal account on the same product. Using a personal account for work files is one of the most common ways confidential information leaks, and it can breach your contract.

    Check your confidentiality obligations. A non-disclosure agreement, client contract or professional duty may forbid sharing material with third parties, and an AI service is a third party. Read the clause before you paste.

    Anonymise first. A simple technique solves many cases. Before pasting, replace real details with placeholders, then put them back after.

    Before: "Chan Tai Man at Lakeside Trading owes HK$84,200 on invoice 2231, due 3 March."
    After:  "[Client A] at [Company B] owes [Amount] on [Invoice], due [Date]."

    Ask the AI to work with the placeholders, then replace them in the result yourself. For a document, remove or mask names, numbers, addresses and anything that would identify a person or a deal.

    Keep sensitive work local or off AI. For truly sensitive material, consider a tool that runs on your own device, or do the work by hand. The local-model tutorial in this track covers running models on your laptop, which can keep data on your machine.

    Keep a record. For client work, note which tool you used for what. If a question comes up later about how a document was produced, you will have an answer.

    Check the output as you would a junior's work. Wrong facts and invented sources are a risk to you and your client. The hallucination tutorial's checking routine applies.

    Step 5 — Copyright and ownership, in plain terms

    Copyright and AI is one of the most unsettled areas of law. Here is what is reasonably clear, what is not and what to do about it. This is general information, not legal advice, and it differs between countries.

    Your inputs. Copyright applies to what you paste in. If you paste a whole book chapter, an article, a paid report or another person's artwork, you may be copying material you have no right to use, and the AI's output may reproduce or closely follow it. A practical rule: only put in material you own, have permission to use, or that is clearly free to use, and prefer a short excerpt for analysis over pasting a whole work.

    The output. Who owns what an AI generates depends on where you are, and the rules are not settled. In several countries, including the United States, copyright generally requires human authorship, and material produced purely by an AI with little human creative input may not be protected. Some places have special rules for computer-generated works. The practical result: you may be able to use the output freely, but you may not be able to stop others from copying it, and a client who pays for it may expect rights you cannot guarantee. Where you add real creative work of your own, such as writing, selection, editing and arrangement, you are in a stronger position. If it matters commercially, check the rules in your country and your client's, and read your tool's terms on how outputs may be used.

    Tools differ. Providers' terms say different things about who may use outputs and for what. Read the relevant part of your tool's terms before using output in something you sell.

    Training on copyrighted works. Whether training AI on copyrighted books, images and articles is lawful is being decided in courts and legislatures around the world. The year-in-review tutorial notes a reported settlement of about $1.5 billion in a US case about books, and that an earlier ruling found training on lawfully obtained books to be fair use while keeping pirated copies was not. That is one district court's ruling in one country, and other cases were still pending. For you, the takeaway is modest: the law is moving, and the safest practice is to avoid asking AI to copy a named living artist's style to pass off as theirs or to reproduce protected text.

    Avoid close copies and pass-offs. Do not use AI to produce something that closely imitates a particular person's work and present it as yours or theirs. That can raise copyright, trademark and honesty problems.

    Respect licences. If you use AI on material under a licence, such as stock images, a course, a template or open-source code, check that the licence allows it. Open-source licences, in particular, can carry conditions.

    Keep evidence of your contribution. If you want to claim authorship, keep your drafts, prompts and edits. They show what you added.


    Step 6 — Other people: faces, voices, names and consent

    AI makes it easy to copy a person. That convenience raises ethical and sometimes legal questions that are about them, not you.

    Faces and likeness. Do not make images or videos of a real person doing or saying things they did not do, unless it is clearly satire and clearly labelled, and never in a way that could embarrass, harass or mislead. The EU has agreed to ban AI systems that create non-consensual intimate imagery, with a compliance deadline of 2 December 2026 in the rules adopted this year, and many countries are moving the same way. Never create or share sexual images of a real person without consent, and never of anyone under 18, in any form.

    Voices. A short audio sample can be enough to clone a voice. Do not clone someone's voice without clear permission. Be aware that the same technology is used in scams, such as a phone call that sounds like a relative or a boss asking for urgent money.

    Names and claims. An AI can write something false and damaging about a real person with total confidence. Do not publish a claim about a named person that came from an AI without checking it against a reliable source. False statements about real people can be defamatory.

    Meetings and recordings. Rules on recording, and on using AI notetakers on calls, vary. Tell people before an AI joins a conversation, and check the law and your organisation's policy about consent.

    Children and students. Be extra careful with anything involving minors, and follow the rules of schools and platforms. Do not put children's names, photos or details into general AI tools.

    Scams. AI makes scams more convincing: fake voices, perfect emails and fake documents. Habits that help: be suspicious of urgency and secrecy, verify any request for money or credentials by calling back on a number you already know, and agree a family or team code word for emergencies.

    Consent as a test. For anything involving another person, ask: would they be happy to know I did this? If not, think again.

    Step 7 — Honesty about AI use, and your one-page policy

    Being open about AI use builds trust, and in some places it is required.

    Say when it matters. If AI did a substantial part of the work, and the reader would reasonably want to know, say so. This is especially true in education, journalism, research, professional advice, product reviews and anything where people rely on your expertise.

    Follow the rules where they exist. Schools and universities have academic integrity policies on AI use, and some platforms require you to label AI-generated or altered content. Regulation is also arriving. The EU's transparency duties, for example, include labelling certain AI-generated content, so if you publish in or for that market, check the official text. Read the rules of any platform you publish on.

    Own the result. You are responsible for what you publish, whoever or whatever drafted it. "The AI wrote it" is not a defence for an error, a copied passage or a harmful statement.

    Do not use AI to deceive. Fake reviews, fake testimonials, fake identities and passing off AI work as your own where that is forbidden are all things to avoid. They harm others and can harm you.

    Write your policy. A short personal policy removes decisions from the moment. Here is a template to adapt.

    My AI use policy (date: ___)
    
    1. I never paste: passwords or codes, financial or ID numbers, other people's private records, material under confidentiality, or unreleased work material.
    2. I trim and anonymise: client names, numbers and anything identifying, before I use AI on work.
    3. I use only tools my employer approves for work files, and I keep personal accounts for personal use.
    4. I checked the privacy settings of my main tools on (date): training off / history on or off / connections reviewed.
    5. I only put in material I own or have permission to use. I do not ask AI to imitate a named living artist or to reproduce protected text.
    6. I do not make images, voices or claims about real people without their consent, and I verify money requests by calling back.
    7. I say so when AI did a substantial part of work that others rely on, and I check every fact before I publish.
    8. I review this policy on (date, three months from now).

    Worked Example: Dara handles a client contract

    Dara is a freelance consultant. A client sends a draft services contract and asks for her view. She is short of time, and the obvious move is to paste the whole contract into her favourite AI and ask for a summary of the risks.

    She stops and runs the three questions. What am I sharing? A client contract, with the client's name, fees and terms, probably under a confidentiality clause. Who can see or keep it? The AI company, under a personal plan whose data settings she has never checked. What could go wrong? A breach of confidentiality, and a client who finds out.

    She applies the traffic-light rule: this is amber, leaning red. She checks the contract and finds a confidentiality clause that forbids sharing with third parties without consent. She chooses another path.

    She replaces the names, fee amounts and dates with placeholders, and removes the clause numbers that would identify the deal. She then pastes only the three clauses she is unsure about, in anonymised form, and asks the AI to explain them in plain English and list questions she should ask. She keeps the risk judgement for herself, and for anything with real legal weight, she tells the client she would like their lawyer to review it.

    The AI helps her understand a limitation of liability clause that she had skimmed. She checks the explanation against the clause text, not just the AI's summary. She writes two lines in her notes: which tool, what was shared in anonymised form, and the date. And after the job, she spends five minutes checking the training and history settings on her account, and turns off the option that allows her chats to be used to improve the product. The whole thing costs her ten extra minutes, and keeps her inside her contract.

    Common Mistakes to Avoid

    Three habits cause most of the trouble.

    Mistake #1: Pasting first, thinking later. A confidential file, a customer list or a private message goes into a chat because it is quick. Fix: run the three questions, use the traffic-light rule, replace sensitive details with placeholders and use only approved tools for work material.

    Mistake #2: Assuming you own, and may use, everything the AI produces. Output may be unprotected, may resemble someone else's work and may come with terms you have not read. Fix: put in only material you have the right to use, add real human work of your own, read the tool's terms on output use and keep a record of your contribution.

    Mistake #3: Forgetting the people who are not in the chat. Faces, voices, names and data belong to people who did not agree. Fix: get consent, avoid false or intimate content about real people, verify any request for money by a second channel and ask whether the person would be happy to know.

    Going Further

    Run a quarterly privacy check. Every three months, repeat the five-minute check of each tool's settings, connections and memory. Settings and defaults change.

    Create approved prompts and placeholders. For recurring work, keep a template that already has placeholders for names and numbers. It makes the safe way the quick way.

    Talk to your team. If you manage others, write down a short, practical AI rule and discuss it. People follow a simple rule that they helped shape.

    Read the official sources. For your country, find the data protection authority's guidance on AI and your professional body's guidance, if you have one. They are written for people in your position.

    What to learn next. The next tutorial is a weekly routine for keeping up with AI without burning out, which includes a habit for noticing when a privacy setting or a rule has changed.

    Key Takeaways

    In this tutorial you learned:

  • Three questions before you paste. What am I sharing, who can see or keep it, and what could go wrong?
  • Use a traffic-light rule. Red is never pasted (credentials, ID and financial numbers, other people's private records, confidential and unreleased material), amber is trimmed and checked, and green is fine.
  • Audit each tool's settings in five minutes. Training, history, deletion, memory, connections, sharing, plan differences and extensions. Note the date.
  • Work material needs work rules. Find the policy, use approved tools, check confidentiality duties and anonymise with placeholders.
  • Copyright is unsettled. Put in only what you may use, expect that pure AI output may not be protected, read the terms and add real human work and keep evidence of it.
  • Other people are not in the chat. Get consent for faces and voices, avoid intimate or false content about real people, verify money requests by a second channel and watch for scams.
  • Be honest about AI use. Say so where it matters or is required, own the result and never use AI to deceive.
  • Keep a one-page policy, and review it every three months.
  • Before you move on, a challenge. Do the five-minute settings check on the AI tool you use most, then write your one-page policy with a review date three months away. Finally, take one task you did recently and ask whether it would have passed your three questions. If not, write down what you would change.

    Learn AI, after work

    Track your progress, earn XP, and unlock more free tutorials in the AfterWork Academy app.

    Open this tutorial in the app

    More AI tutorials