Tech for Beginners Part 43: Using AI Safely: Privacy, Wrong Answers and Deepfakes

AI tools can explain, write, translate and plan in seconds, which is why millions of Indians now use them every day. But the same technology brings new risks. People paste private documents into chatbots without thinking. AI answers that sound perfectly confident turn out to be wrong. And scammers now use AI to copy voices and faces, making a fake call from your "son" or a video of a "famous businessman" recommending an investment look real. This is Part 43 of our Tech for Beginners series. Today we will learn what you should never paste into a chatbot, how to double-check AI answers, how AI-generated voice and video scams work, and how to spot deepfakes and verify before you believe.

Many AI scams are new versions of old tricks. The fake police video calls in Part 26 on digital arrest scams and the investment traps in Part 25 now sometimes use AI-made voices and videos. If you are targeted, the reporting steps in Part 28 (1930, cybercrime.gov.in and Sanchar Saathi) still apply.

A note on changes: AI is improving quickly, and deepfakes are getting harder to spot by eye. The warning signs below help today, but the most reliable protection is the habit of verifying through a separate, trusted channel. Privacy settings in AI apps also change, so check each app's current settings.

Part 1: Protect your privacy with chatbots

What happens to what you type?

When you use an AI chatbot, your messages are sent to the company's computers. Depending on the service and your settings, your conversations may be stored, reviewed by people to improve the service, or used to train future AI models. Even if a company protects data well, it is wise to assume that anything you type could be seen by someone else.

Never paste these into a chatbot

• Aadhaar, PAN, passport, voter ID or driving licence numbers, or photos of these documents.
• Bank account numbers, card numbers, CVVs, UPI PINs, passwords and OTPs.
• Full medical reports with your name and details. Remove identifying details first.
• Office documents marked confidential, customer lists, or company plans, unless your employer has approved the tool.
• Other people's private information, such as a friend's phone number, address or private messages.
• Private photos of yourself or your family that you would not want to be shared.

How to share safely

1. Replace personal details with placeholders: [Name], [Account number], [Address].
2. Copy only the paragraph you need explained, not the whole document.
3. Crop photos to remove names, numbers and faces.
4. Check the app's privacy settings and turn off the option to use your chats for training if you prefer.
5. Delete conversations you no longer need.

Beware of fake AI apps and sites

Scammers create apps and websites with names like "AI Chat Pro" or fake versions of famous AI brands. They may charge large subscriptions, show endless ads or steal data. Install AI apps only from the official Play Store or App Store listing and check the developer: OpenAI for ChatGPT, Google LLC for Gemini, Microsoft for Copilot. Never install AI tools as APK files sent on WhatsApp or Telegram.

Part 2: Double-check AI answers

AI chatbots can produce false information that sounds convincing. This is sometimes called a hallucination. They may invent:

• Laws, rules or court judgements that do not exist.
• Government schemes, eligibility criteria or deadlines.
• Statistics, quotes from famous people, or book titles.
• Website links, phone numbers and email addresses, which can even lead to scam pages.
• Wrong calculations, such as loan EMIs or tax amounts.

A simple checking routine

1. Ask for sources: "What is your source? Give me the official website." Then open the source yourself and confirm it says the same thing.
2. Use official websites for anything about government, banks, taxes, exams and laws: sites ending in .gov.in, .nic.in, or the organisation's own website.
3. Never call a phone number provided by AI for a bank or company helpline without confirming it on the official website or the back of your card.
4. Recalculate numbers with a calculator or an official online calculator.
5. Ask a second assistant or do a normal web search. If answers differ, dig deeper.
6. Consult a professional for health, legal and financial decisions.

Part 3: AI voice and video scams

Voice cloning

With just a short recording of someone's voice, perhaps from a video posted online, AI tools can create a fake voice that sounds like them. Scammers use this to:

• Call parents pretending to be their son or daughter "in trouble", asking for urgent money.
• Pretend to be a boss asking an employee to make an urgent payment.
• Pretend to be a relative abroad who needs money for an emergency.

Deepfake videos

A deepfake is a video or image where AI has been used to make a person appear to say or do something they never did. In India, deepfakes have been used to:

• Show famous business leaders, cricketers or film stars "recommending" investment apps or trading groups.
• Show fake doctors promoting miracle medicines.
• Create fake video calls from "police officers" or "officials".
• Create fake obscene images of real people for blackmail or harassment.

Part 4: How to spot a deepfake

These signs can help, though good deepfakes may show none of them:

In videos

• Lip movements that do not quite match the words.
• Unnatural blinking, stiff facial expressions or a face that looks "pasted on".
• Blurry or flickering edges around the face, hair, glasses or earrings.
• Skin that looks too smooth, or lighting on the face that does not match the background.
• Strange hands, teeth or jewellery.

In voices

• A flat, robotic or oddly even tone, or unusual pauses.
• Background noise that suddenly changes or cuts out.
• The caller avoids answering personal questions or keeps the call very short.

In the situation itself

This is the most reliable clue. Ask:

• Is there urgency and pressure to act now?
• Am I being asked for money, OTPs or secrecy?
• Is a famous person promising guaranteed returns?
• Is this coming from an unknown number or a new account?

If the answer to any of these is yes, treat it as a scam until proven otherwise.

Part 5: Verify before you believe

1. Hang up and call back on the number you already have saved for that person, not the one that called you.
2. Set a family code word: agree on a secret word or question with close family. In a real emergency, the caller can say it. A scammer cannot.
3. Ask a personal question that only the real person would know, and that is not on social media.
4. Check official channels: a real celebrity endorsement or government announcement will appear on official, verified websites or accounts and be covered by reliable news organisations.
5. Search for fact-checks, as explained in Part 38, using a description of the video plus the words "fake" or "deepfake".
6. Take your time. No genuine emergency is made worse by a two-minute call to confirm.

Reduce your own risk

• Keep social media accounts private, so fewer strangers can copy your voice, face and family details.
• Think twice before posting long videos of yourself speaking, or children's photos publicly.
• Do not answer video calls from unknown numbers.
• Turn on WhatsApp's Silence unknown callers setting, covered in Part 36.

Teach the family

Older parents and young children are often the most at risk. Have a short family talk:

• Explain that voices and videos can now be faked, so "I heard his voice" is no longer proof.
• Practise the code word and the "hang up and call back" rule.
• Show them one example of a deepfake news story from a fact-check website.
• Tell children never to share private photos online, and to tell you immediately if anyone threatens them.

If you are targeted

• Money lost: call 1930 immediately and file a complaint on cybercrime.gov.in.
• Fake or obscene images of you: report the content to the platform, save the links and screenshots as evidence, and file a complaint on cybercrime.gov.in. Do not pay blackmailers.
• Deepfake ads using famous people: report the ad or post on the platform where you saw it.
• Suspicious calls: report them through the Chakshu facility on Sanchar Saathi.

Frequently asked questions

Can AI tools tell if something is a deepfake?

Some detection tools exist, but none is perfect, and results can be wrong. Do not rely on them alone; verify through trusted people and official sources.

Is it safe to upload my photo to AI apps that make cartoons or old-age pictures?

Be careful. Read what the app does with your photos, use only well-known apps, and avoid uploading clear photos of children. Your face is personal data.

Is using AI to write my office work wrong?

It depends on your employer's rules. Many workplaces allow AI help but forbid sharing confidential data. Ask your manager or check company policy.

What is next

Try this today: agree on a family code word for emergencies, and check the privacy settings in the AI app you use most. In the next part of Tech for Beginners, we start a new section on computers with a beginner's checklist for buying your first laptop or desktop.