Audio and Video Can Be Faked: A 10-Minute Protocol Before You Trust or Share

If a video features a familiar face, a recognizable voice, and a message that feels urgent, we tend to believe it right away. The problem is, this combination can now be created or manipulated with the help of artificial intelligence (AI). Voices can be mimicked, faces can be superimposed onto other videos, and old footage can be given new context.

The Ministry of Communication and Digital reminds us that deepfakes—audio, images, or videos manipulated with AI technology—are becoming increasingly difficult to distinguish from original content. In April 2026, Komdigi also denied a video that misused the account of the Secretariat General of the Ministry of Finance, urging the public to provide data to receive assistance. The audio in the video was strongly indicated to be the result of AI manipulation. Read Komdigi's explanation regarding this case.

This means we can no longer rely on the phrase "I saw the video myself" as the sole reason to believe. Videos are clues, not always final proof.

Don't start with the question "Is this real or fake?"

This question is indeed important, but it is often difficult to answer just by looking at it briefly. High-quality videos can appear natural, while original videos can also be edited, misleadingly captioned, or shared out of context.

A more useful question is: who is the original source, what is the context, and what action is being requested from me? These three questions help us examine the content without needing to be a digital forensics expert.

For example, a video of someone supposedly announcing a relief program needs to be verified through the official account of that institution. If the message in the video requests identification numbers, OTP codes, administrative fees, or transfers to personal accounts, the risks are much greater—even if the face and voice appear convincing.

10-Minute Protocol Before Trusting

1. Pause Before Reacting

Digital scams often work by exploiting emotions: panic, fear of missing out, anger, or excessive excitement. Messages like "send now", "don't tell anyone", or "this opportunity is only available today" should be a signal to slow down your decision-making.

Do not immediately forward the video, click on links, or contact the numbers listed in it. Save it first, then check separately.

2. Look for the Earliest Source

Check who first uploaded the content. Accounts with names similar to official institutions are not necessarily official accounts. Look at the account address, upload history, verification badges if available, and whether the account has a consistent communication pattern.

For statements from officials, aid promotions, or public service announcements, manually visit the official websites and accounts of the institutions. Do not rely solely on links sent through private messages.

3. Compare with Independent Sources

Important content usually does not appear on just one account. Search for key terms from the video, then compare with credible media coverage or fact-checking channels. If a major claim is only circulating through one anonymous account, the level of caution should be increased.

CISA, the U.S. government's cybersecurity agency, advises the public to use reverse image searches and check media sources to help find the original context. This technique is useful for determining whether an image or video clip actually comes from an old event. See CISA's guide on disinformation tactics.

4. Focus on Content, Not Just Technical Quality

Many people look for signs of deepfakes on faces: strange lip movements, unnatural blinking, or inconsistent lighting. These signs can sometimes help, but they are not always visible. Technology changes, and videos compressed by messaging apps can also appear distorted.

It is safer to check for contextual inconsistencies, such as:

Accents or speaking styles that differ from the habits of the person shown.
Movements of the mouth that do not always match the voice, but the video quality is too low to confirm.
Logos, titles, dates, or institution names that appear inconsistent.
The video is cut off just before important parts that should explain the context.
Messages requesting financial actions or personal data without official procedures.

5. Verify Through Known Channels

If you receive a call or voice message claiming to be from a family member, friend, boss, or business associate, do not verify through the number provided in the message. Contact a number already saved in your phone or use another communication channel.

The Federal Trade Commission warns that scammers can use short recordings available online to mimic the voices of family members. The practical advice is simple: do not trust voices alone, contact the person through a known number, and do not send money just because there is an urgent request. Read the FTC guide on voice cloning scams.

Add a "Safe Word" in the Family

For families, especially those who often receive calls from unknown numbers, create a verification word or question that is not easy to guess. For example, it should not be a birth date or pet's name, but rather a specific phrase known only to family members.

A safe word is not a substitute for full verification. However, in panic situations, a simple question can help distinguish a genuinely known person from a scammer who only has voice samples or information from social media.

What Does This Mean for Us?

We do not need to recognize every AI technique to stay safe. The most important habit is to separate identity from requests. Someone may look exactly like a person we know, but a request to send money must still be verified through other channels.

The same applies to videos of officials, celebrities, experts, or public figures. The reputation of the person appearing is not proof that the video is authentic. Check their official channels, look for reports from other sources, and question whether the claims make sense.

What You Can Do Now

Stop the habit of forwarding messages just because they are sent by someone you know.
Save the official numbers of banks, family, schools, offices, and important institutions separately.
Enable two-factor authentication on important accounts.
Limit public uploads that contain long audio, personal data, or routine location information.
If you encounter suspected fraud, do not share it again even if the intention is to warn. Save the evidence and report it through official channels.

In the midst of the proliferation of synthetic content, being skeptical does not mean always being suspicious of everyone. Skepticism means taking a pause before trusting, sharing, or taking actions that are hard to reverse. Ten minutes to check can be much cheaper than regretting after money, data, or reputation is already lost.

Sources & Further Reading

– Rio Yotto @rioyotto