When deepfakes erase the line between truth and falsehood, how can we verify our identity?

AI and deepfakes are blurring the lines between "real" and "fake," even questioning human identity. Trustworthiness is becoming increasingly difficult to prove, while suspicion is easy. A crucial question, therefore, is not just how we can avoid being deceived, but how we can prove that we ourselves are "authentic."

Current AI technology is developing at such a rapid pace that the line between... "The real thing" with "Things that were created" The situation has become more blurry than ever before. Previously, most people were concerned about being scammed by fake information, but now it's not easy to even tell the person on the other end of the line that it's really you.

In particular, the use of AI for deepfakes and AI-powered voice cloning is a major issue that can pose systemic risks to social and economic trust, as well as to daily life. Because building trust becomes more costly, while questioning it is incredibly easy, another question that should be asked might not just be, "How can we avoid being deceived?" but rather, "..." "In a world where everything is easily questioned, how can we prove that we are truly who we are?"

This is the situation Israeli Prime Minister Benjamin Netanyahu faced last March. It started with the posting of a video that contained a reflection that made it appear as if Netanyahu had a sixth finger on his right hand. This was initially a clear indication that the video was a deepfake. This led to online rumors that he had died and that Israel was covering up the truth.

But a few days later, Netanyahu posted another video of himself in a coffee shop, raising his hand and smiling to show that he had all his fingers like a normal person.

What does this reflect?…This is the first time a national leader has had to openly attempt to prove that he is not an AI, and it seems to have failed, as many people still believe Netanyahu is dead. Although video experts have confirmed that Netanyahu's video is authentic, they point out that the anomalies seen are common in videos. For example, the sixth finger is simply a reflection on Netanyahu's palm, which might look strange when the footage is paused at that moment, but there's nothing more to it. In fact, AI technology hasn't created extra fingers in years, and a model capable of producing video of this caliber shouldn't miss this detail.

Image from youtube.com/@IsraeliPM/videos

Another piece of evidence confirming that Netanyahu's six-inch video is not a deepfake is that there's a moment where he accidentally bumps into the microphone, causing background noise that cuts out his speech. This kind of continuity is very difficult to replicate. Meanwhile, Netanyahu's video in the coffee shop is also authentic.

Hany Farid, a professor of digital forensics at the University of California, Berkeley, and co-founder of GetReal Security. He stated that his team analyzed the audio, detected faces frame by frame, and meticulously examined lighting and shadows. "There is no evidence to suggest that this video was created by AI."

But all that wasn't enough. Netanyahu even posted a third video, while disbelievers remained unbelieving. That raises a question that should be asked in return: "How can we prove that it's really us?"

Thomas Jermaine, Senior Technology Correspondent for the BBC. He posed this question, and he actually began the experiment by video calling his aunt and saying: "I'll video call you again soon so you can see if you're talking to a real person or an AI." Because he wanted to prove whether people who had known him his whole life could tell the difference. He didn't use AI, but simply made another video call.

At the beginning of the experiment, Jermaine's aunt, unaware of the situation, said that... "The voice sounds like a real person, with tonal variations that sound more like a real person than AI." But when Jermaine said that AI is now very advanced, and that those voice tones could be easily replicated, his aunt became hesitant and said... "At first, I was 90% sure it was real, but now I'm starting to doubt it because it started looking more like a fake."

The hesitation caused the truth to begin fading. Jermaine's aunt tried reading jokes from Facebook to see if his reactions were natural, and Jermaine laughed normally. While this helped her feel better, she remained unsure. Even minor deviations from his usual behavior, like changing the color of his favorite shirt, led to speculation that he might be an AI. Eventually, Jermaine revealed he was real, but his aunt still felt uncertain.

This experiment shows that even people who have been close to someone their entire lives can be hesitant about whether it's a real person or an AI. Even slight deviations from the norm can easily lead to it being perceived as AI. This becomes much more difficult when the person involved is someone you don't know well, especially when it's a prime minister and the audience is a global audience.

Jermaine also interviewed experts about what method could immediately prove to people that he/she is not an AI, to which the answer was very short: "There is none." However, there are methods to test and reduce the possibility that it's not AI, such as checking the ambient sounds, shadows behind the subject that correspond to movement, reflections on glasses, or observing other activities like picking up a book or taking notes during a conversation—things a deepfake wouldn't do. But these methods aren't truly guaranteed; they only reduce the likelihood of AI. Without these steps, it's almost impossible to be certain whether the image is of a real person or an AI. Even those technologically savvy find it very difficult to confirm whether the other party is real.

In the early stages of the Ukrainian war following the Russian invasion in 2022, deepfakes seemed unreliable. However, over time, particularly during the Gaza conflict, the quality of fake content improved significantly. By the time it reached Venezuela, fake content had increased rapidly, and even Iran took it to another level.

Experts told Jermaine that in Netanyahu's case, the problem was that the team used high-quality cameras and set a depth of field with background blur, a technique similar to AI-powered video, which raised suspicions.

Furthermore, today's world is full of fake content, making distrust commonplace and difficult to eradicate.

The approach proposed by world-class experts is not complicated; in fact, it's so simple that even previous generations could have conceived it. It involves using a "secret code." Each person, family member, business partner, and those needing to discuss important matters should have a unique, known only to them, for authentication in emergencies, similar to multi-factor authentication.

It's important to know that deepfakes are becoming a primary tool for scammers. According to the American Association of Retired Persons (AARP), AI-powered scams increased 20-fold between 2023 and 2025, with victims ranging from individuals to large organizations. A British engineering firm even lost over $25 million after being scammed by a deepfake of its CFO, and this trend continues to rise.

While the straightforward answer is, "There's no way to 100% confirm from a conversation that the person on the other end isn't an AI," in practice, we can mitigate the risk using the following methods.

1. Observe behaviors that AI still can't imitate convincingly: Even though AI is very advanced, there are still some points that can be observed, such as:

• Continuity of context: A truly resilient person will remember previous conversations with specific details. Try asking about particular topics you've discussed before.
• Imperfect reaction: Real people hesitate, stammer, and change words mid-sentences. AI is often too smooth or too good to be true.
• Answering questions off-script: If you ask a sudden or very specific question, AI will sometimes give a broad answer or avoid the topic.

2. Use the secret code: This is the method actually recommended by experts, and it's the simplest and most effective: Establish a secret code with a close friend or business partner to use for identity verification in important situations, such as financial transactions or urgent decision-making. This acts like two-factor authentication (2FA) in the real world.

3. Confirm via another channel: Don't trust only one source. If someone calls, try hanging up and calling the same number back. If you receive a strange message, double-check through another app. Always double-check, especially when it comes to money matters.

4. Pay attention to visual/auditory details: During a video call, pay close attention to details such as audio and video synchronization, lighting and shadow continuity, movement, and whether there is genuine interaction, such as picking up or dropping an object. However, it's important to understand that advanced AI is already capable of performing these tasks.

5. Be aware of unusually urgent situations: Deepfakes are typically used in situations requiring urgent decisions, such as immediate money transfers, confidentiality measures, or extreme urgency. The risk is particularly high when dealing with pressure, confidentiality, or financial matters.

AI technology has advanced significantly. Simply relying on conversation alone is no longer sufficient to be 100% certain you're not being scammed. The best approach is to use a combination of methods: secret codes, careful observation, and finally, no matter how urgent or important something is, never believe it immediately, especially when it involves money.

By the editorial team bank finance

refer : bbc.com, bloomberg.com

Read news related to Technology circles around the world can be found here.