A Video is Shared in the Family Group Chat
It is 10 p.m. Your family group has a clip: Someone you know, a celebrity or a family member is saying something crazy or a voice note from a family member asking for funds following an accident. Five family members have already responded in a shocked manner. One has passed it on to another group. You're going to do the same.
That is where deepfake verification comes in and praising the good Lord is not primarily an eye test. It is a test of habits. Everyone can be deceived and it is not a bad trait. This guide provides you with a 5 step process to follow in a group chat, the reasons why detection tools alone are not the solution, and where to report harmful content in India.
The reasons why your eyes and ears are not enough
It's easy to imagine that you will simply see the flaws. But there is evidence otherwise. Conducted by identity company iProov, the study in 2025 revealed that 2,000 UK and US participants were presented with a combination of real and fake images and videos, and only 0.1 per cent of them correctly identified all the items, despite being instructed to look for 'deepfakes'. 36% less were able to determine a fake video than a fake image, and more than 60% were too certain in their own skills. It's also noted that just 11 percent critically analyzed the source and context of what they saw, and 18 percent did not bother to analyze any of the information.
Handle with caution: it is a vendor study, not an Indian study and it was an artificial quiz. It does seem to have a lesson for everyone, though. It's hard to detect visual glitches and the one thing that most people will miss is checking the source and context.
The Five-Step Verification Routine
Step 1: Pause sharing
The strongest action is the simplest: Do not forward it just yet. All reserves are distributed, even with the addition of "is this real??" The successful deep fake is one that spreads faster than anything else, and there's nothing more compelling than emotional material to get you moving quickly.
Apply the 60 second rule to any shocking, frightening or money-requiring thing.
Notice your reaction. Anger, panic and outrage are the indicators that it is time to slow down, not speed up.
Always answer with a simple phrase: "Checking this before passing on. That also allows other people to take a break.
Step 2: Find the text in context.
Point out the source of this. If a clip is cut off from its source, it's a red flag. Look for the latest version you can.
Check the source of posting, date & location and if the account has a history.
Capture a key frame image and perform a reverse image search. There are some tools that allow you to do this in seconds.
Check for an extended version. The clips used are often manipulated and may be short and cut.
What was not in the story: time, place, original speaker, event name.
This is important because the deepfakes and miscaptioned actual videos are sent without backstories. Remember that metadata is usually lost in sharing files on chat apps, so no details are conclusive.
Step 3: Cross-check credible sources
When something of significant importance or a statement is true, it will be reported by reliable sources. Be wary of anything that is only forwarded, not included in the original clip.
Look for claims on trusted news sites and the person or organization's official sites.
Look at trusted fact checking websites, both domestic and international, such as PIB Fact Check in India and BOOM.
Do not use it as a source "I think lots of people in my group are sharing it.
If a clip states that someone said or did something, research the clip to determine where and when it was said.
Step 4: To carefully examine the claim.
Next, observe the expectation the speaker is putting on you in the clip: what is he asking you to believe or do? Keep the claim away from the digipixels.
Does the assertion stand out from common knowledge? Is it similar to what someone has said or done in the past?
Does it require authenticity, discretion or funds? These are all tell-tale signs of fraud.
Look for odd pacing, flat affect or audio that is not appropriate for the situation, and listen for odd audio and lip movement, strange lighting, blurred edges or distorted hands. These are hints and not proof. There is such a thing as a fake clean clip and a glitchy clip can be the real thing.
Use a family safe word, if the issue is money, or an emergency, check with the person directly by calling a number you already know. Do not use a number in the message.
The idea here is to consider the claim along with the source rather than simply watching the video until you make a decision.
Step 5: Report harmful content
If it is a fake, harmful or fraudulent, report it. DON'T just delete it and forget it exists as others may still see it.
Save Evidence First: Screenshot of the post, the account, and link before it gets taken off the internet.
Report on the Platform: Where you saw it, using its tools for manipulated media, impersonation or harassment. In India file a complaint on the National Cyber Crime Reporting Portal (cybercrime.gov.in). If money is involved, call the cyber-fraud helpline 1930 immediately.
Be familiar with platform Duties: The amendment to the IT Rules, notified on 10 February 2026, introduces the term ‘synthetically generated information’; introduces obligation to tag and the three hour due process to remove illegal AI content; and extends to in-scope platforms.
For Intimate Imagery: There are services available to help remove intimate imagery, such as StopNCII.org (for adults) and Take It Down (images of minors). Check their process first to ensure that it is correct.
Correct, do NOT amplify: In the group, share the verified correction or fact check, NOT the fake.

Realistic Scenarios
The ideas presented in these examples have been created from patterns and are hypothetical.
The Emergency voice note. The message is a voice note from a cousin that has been received by the family group that he has had an accident and needs cash to a number immediately. Do the steps: stop, dial your cousin's saved number, and inquire about a family safe word. Take him up, if he picks it up and is okay, you're right.
The political clip. A video is running around the neighbourhood group the night before an event where a politician is seen saying something inflammatory. The running steps: pause, search for the whole speech/factual coverage, look at a fact-check site, post the verified coverage rather than the clip.
The video is called "Boss". The accounts employee gets a video call from a man who sounds and looks like the finance director, requesting a confidential transfer of funds. This is not a figment of the imagination: there's a widely reported example of a £20m deepfake attack on engineering firm Arup where criminals used deep fake video to assume the identities of executives. Perform the following steps: end call, confirm on an independent trusted channel, adhere to dual approval procedures for payments.
The Limits of Detection Tools
Automated detectors are useful, but they are not a verdict. Here's why you need to consider them just one clue among a few. They have difficulties with new fakes. For Facebook's Deepfake Detection Challenge, the best model achieved an 82.56 percent accuracy on the public data set, and a 65.18 percent accuracy on a separate test set of unseen videos. That's in 2020 and tools are better, but in reality it's still the gap between a lab benchmark and the real world. The challenge's organizers actually decided that context and provenance are more important than analysing the pixels.
The evidence is compromised due to chat apps. When data is compressed and re-shared, those subtle signals that detectors depend on are lost.
Both mistakes happen. A detector can be inaccurate and misclassify a legitimate clip as fake, or be correct and misclassify a fake clip as legitimate, and a high confidence score does not mean it's a correct score.
No Guaranteed Labels. India has mandated that platforms label synthetic content and retain its provenance markers, with the obligation to label synthetic content falling on in-scope platforms. Legal interpretation of the rules suggests that content can't be considered "text-only," and an unlabelled clip does not constitute proof of the clip being genuine.
Attackers adapt. If any method of detection comes to light, it can be attacked.
The Thumb Rule: "the detector says real" does not mean "the detector says give", and "the detector says fake" does not mean "the detector says prove". Follow steps 1-4 with it.
Group Chat Habits That Prevent Harm
Share the five steps with your oldest and youngest family members, they might be most vulnerable and least questioning. Put a note in the group, for instance: "We check before we forward. Reward the person who stops, do not shame the person who was fooled, shame will prevent them from reporting. In the case of community or workplace groups, include a rule that requests for money of an urgent nature are always confirmed with a second means.
Quick-Reference Card
1. Pause: Do not forward for 60 seconds.
2. Locate: Find and identify the original and context.
3. Cross check: Find credible and independent reporting.
4. Inspect: make judgment about the claim, the urgency and the ask and not the pixels.
5. Report: Save evidence and report it and take part in sharing the correction.
For Organizations: Where Thread Security Fits
Deepfakes are a threat for businesses as executives may impersonate themselves or customers to get money and reset passwords. According to the 2026 India report released by IBM, 26 percent of the malicious attacks in India were AI-driven. Thread Security calls itself a “certified offensive security company & penetration testing firm” and its listed services include “web application penetration testing”, “cloud security reviews”, and “smart contract auditing. The tests assess the systems that an impersonator attempts to access, including login, reset and payment flows. IBM further revealed that offensive security testing was the biggest cost savings factor, saving approximately ₹2.47 crore per breach. Social engineering assessments and staff awareness work are largely a part of deepfake verification, so check with any vendor, including Thread Security, if they offer them or not.
Cyber Security Courses in Jalandhar: How to do Digital Unmasking!
Verification skills are used daily by security operations analysts, fraud investigators and incident responders. For those searching for cyber security courses in Jalandhar, Thread Security Education provides courses with offline classes and live batches. Their website states that they offer a Cyber Security Course (CEH and industrial training), Defensive Cyber Security with Blue Team and SOC operations, Web Penetration Testing and Bug Bounty Hunting courses, and AI tracks such as AI Security and Responsible AI Engineering. Look to see if any module has digital forensics or deep fake analysis as they were not listed separately.
Frequently Asked Questions
Is it possible to tell if a face is a deepfake or not just by looking? Not reliably. But in one study conducted in the year 2025, just 0.1 percent of people got all of the true and false items right. Look at the source, context and claim, not just the images.
Do deepfake detection tools work? They are there to assist but not evidence. They can cause false positive as well as false negative results, and they can cause performance drop when they are not seen on the fakes or when the file in the chat app is compressed.
What shall I do if I fall victim to a deep fake? Keep evidence, report to the platform, file a complaint at Cybercrime.gov.in, call 1930 in case of monetary transactions, inform a trusted person. If the photo is close-up, inquire about removal services. Consider speaking to a lawyer.
Is it illegal to share a deepfake in India? Depends on the content and purpose. Cheating, impersonation and harassment across synthetic media is a criminal offence, and platforms have new responsibilities in the amended IT Rules of 2026. If for any particular situation, a lawyer is required, then see a qualified lawyer.
Key Takeaways
The claim, not the evidence, is a convincing clip. Deepfakes are not about what you see, but about what you do.
Follow 5-step guidelines: pause, find the original, check for credible sources, check the claim and report harm.
Clues include detection tools. In a major challenge, only a hidden-test score of 65 percent demonstrates the difficulty of detection in reality.
Report Quickly: Save Evidence, utilize the platform, cybercrime.gov.in, and share corrections rather than make-up them up.
Call to Action
Pin up the five-step routine in your family group today and give it to one for-warder who will send it along to another family member. If your organization needs to test your login, reset and payment processes for impersonation, visit threadsecurity.org. Learn investigation and security skills in Jalandhar at Thread Security Education.
Simranjit Kaur
VERIFIEDAI Trainer · Thread Security
Faculty mentor & research guide at Thread Security Education.



