How to Protect Yourself From AI Scams and Deepfake Videos

A more useful first test is: what is this person trying to make me do right now? Anti-scam guidance warns that AI and deepfake scams often create urgency to push people into rushed decisions.

Stop if a message or caller asks you to:

• Send money, make a payment, buy cryptocurrency or change bank details.
• Share a password, one-time passcode, text-message code, ID document, financial information or account screenshot.
• Click an unfamiliar link, scan a QR code, download an app, install a browser extension or open remote-control software.
• Keep the request secret from family, colleagues, a bank, customer support or a supervisor.
• Act because it is supposedly the last chance, an emergency, a rescue situation or an urgent executive request.
• Do something unusual in the name of a celebrity, manager, relative, bank, support agent or government office.

When two ingredients appear together — a trusted identity and pressure to act immediately — leave the original channel and verify elsewhere.

A five-step routine for suspicious messages

1. Pause: stop the interaction

Do not reply, click, download, scan or share. If the pressure is happening on a phone call or video call, hang up. A genuine emergency can survive a few minutes of verification.

2. Switch channels: do not verify through the message itself

Do not use the phone number, link, customer-service portal or account supplied in the suspicious message. Instead, use a number you already saved, an official website you type yourself, an official app, a company directory or another contact method you already trust.

This step matters because if the original message is fake, the phone number, link and support account inside it may be fake too.

3. Use a second confirmation for money and accounts

Families, companies and teams can agree on a simple rule in advance: anything involving payments, loans, bank-account changes, login codes, ID documents, emergency help or remote access requires confirmation through a second channel.

At work, do not approve a payment because of one call, one voice note or one video clip. Deepfake scams can use AI-generated media to impersonate trusted people, so seeing a manager on video or hearing a familiar voice should not replace normal approval controls.

4. Check the original source and full context

For dramatic news, celebrity clips, political videos, disaster updates or investment claims, do not rely on a cropped clip. Deepfakes can manufacture fake video or audio that makes people appear to say or do harmful things.

Ask four questions before sharing or acting:

• Who first published it?
• Is there a full video, full article or original post?
• Do the time, location and event details match?
• Have reliable news outlets, official institutions or the person involved confirmed it?

If you can only find reposts, screenshots or short clips with no original source, do not pass it on yet.

5. Save evidence, then get help or report it

If you think you may have been scammed, preserve the evidence before deleting anything: screenshots, URLs, phone numbers, account names, transaction records, wallet addresses, video files and chat logs. If a bank account, card or payment app is involved, contact the institution through its official channel immediately.

In the United States, a government AI-scam flyer lists Federal Trade Commission reporting options including ftc.gov/complaint and 1-877-FTC-HELP. The FTC also maintains resources on artificial intelligence, consumer advice and scams that can be useful for further reading. Outside the U.S., use the reporting channels provided by your local police, financial institution or official anti-fraud authority.

How to do a quick deepfake check

Visual and audio checks are only a first warning, not a final verdict. Use them as a reason to pause, not as proof.

Look for signs such as:

• Lips that do not quite match the words.
• Flickering or distortion around the face, glasses, hair or ears.
• Odd-looking fingers, teeth, skin texture or facial edges.
• Background text, lighting, shadows or reflections that do not make sense.
• Speech rhythm, pauses or emotional reactions that feel unlike a normal conversation.

But remember the reverse: if you do not see glitches, that does not prove the clip is real. The danger of deepfake scams is that they can look and sound real.

The stronger approach is source verification: find the original publisher, the full clip, the time and place, the surrounding context and confirmation from other reliable sources. For deepfakes, fake videos and misinformation, context usually beats guesswork.

Three everyday rules that make scams harder to pull off

For families: agree on a private check

Families can set a private passphrase or agree that requests for money, emergency help, a new phone number or a verification code always require a call-back. If an older relative receives an urgent message claiming to be from a child, grandchild, police officer, bank or support agent, the safest next step is to stop and ask another trusted family member to help verify it. Deepfake scams can impersonate trusted people.

For workplaces: payments and bank changes need approval controls

If someone claims to be an executive, finance colleague, supplier or client and asks for a payment, account change or login code, follow the established process. Do not skip checks because the voice or video seems familiar. Deepfake content can use fake audio or video to create the impression that someone authorized an action.

For social feeds: the more emotional it feels, the slower you should share

False or manipulated content often spreads because it triggers anger, fear, sympathy or outrage. Be especially careful with posts that say to share immediately, claim that the media will not report it, allege a cover-up, demand instant rescue action or promise guaranteed returns. Pause, check the original source, then decide whether it deserves your attention.

The bottom line: rely on a process, not perfect eyesight

AI makes fake voices, fake images and fake identities easier to insert into everyday messages. Deepfake scams can also use images, video or audio to impersonate trusted people.

What you can control is your routine: do not respond under pressure, do not verify inside the suspicious message, and do not transfer money or share codes just because something sounds or looks like the right person. When in doubt, stop, switch channels and verify.