I tried to prove I wasn’t an artificial intelligence. My aunt wasn’t convinced
Artificial intelligence has reached a point no one expected this soon: it doesn’t just imitate voices and faces with terrifying precision, it’s also destroying something far more valuable than any personal data — our ability to prove we’re real. 😮
Sound like an exaggeration? Then imagine calling someone who’s known you since you were a kid, chatting for a few minutes, and that person still hanging up without being absolutely sure they were actually talking to you.
That’s exactly what happened to BBC journalist Thomas Germain when he ran a simple experiment with his Aunt Eleanor. He called her and explained he’d be making a second call, during which she’d either be talking to the real him or to an AI-generated deepfake. Her job was to figure out which one it was. Eleanor started off confident. She said she thought it was her nephew because the voice had inflections that felt too natural to be machine-generated. But when Germain mentioned the technology was getting pretty advanced, there was a long silence. His aunt backpedaled and admitted it was starting to sound more artificial.
This wasn’t some elaborate deepfake created in a professional studio. It was something anyone with a computer and internet access could pull off today. And if you think this is only a problem for the person on the other end getting fooled, brace yourself: the most disturbing twist in this story is that the threat now comes from the other side.
The problem isn’t just being tricked by a deepfake. The problem is not being able to prove you aren’t one.
When even a prime minister fails to prove he’s real
If a loving aunt already struggles to recognize her own nephew, imagine the challenge when your audience is the entire world. That’s what happened to Israeli Prime Minister Benjamin Netanyahu in an episode that took on surreal proportions on social media.
Netanyahu posted a video where a lighting trick created the impression he had a sixth finger on his right hand. That detail — which a few years ago was a classic telltale sign of AI-generated images — triggered an avalanche of online rumors. People started claiming Netanyahu had died in a missile attack and that Israel was using a deepfake to cover it up. The theory spread quickly and with impressive conviction.
Days later, the prime minister recorded a second video in a coffee shop, smiling and raising his hands to show he had the normal number of fingers. When even that didn’t work, a third video followed. Still, a significant chunk of the public remained convinced everything was fabricated.
According to experts consulted by the BBC, this was the first documented case in which the leader of a major world power openly tried to prove he wasn’t an AI creation. And failed miserably. Jeremy Carrasco, co-founder of the independent publication Riddance, which focuses on AI-generated media, analyzed the videos and quickly concluded they were all real. The supposed sixth finger, he said, was just light reflecting off Netanyahu’s palm. It looks weird if you pause at the right moment, but that’s all it is.
Carrasco also pointed out that extra fingers aren’t a typical AI glitch anymore. The best tools stopped making that mistake years ago, and a model capable of generating the rest of the video at that level of quality simply wouldn’t make that kind of slip. On top of that, at one point in the video, Netanyahu bumps the microphone, producing a sound that interrupts the audio of his speech. That kind of physical and sonic continuity is extremely hard to reproduce with current AI tools.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
Hany Farid, a professor of digital forensics at the University of California, Berkeley, and co-founder of GetReal Security, also confirmed the videos were authentic. His team ran voice analysis, frame-by-frame facial detection, and detailed inspection of light and shadows. The verdict was clear: there was no evidence the content was generated by artificial intelligence.
But for a portion of the public, none of that mattered. Their minds were already made up.
When suspicion hits close to home
Germain himself went through a similar situation in real life, weeks before the experiment with his aunt. He had written an article about a rarely used Google privacy setting and was so excited about it that he shared the link in the family group chat asking everyone to click on it. His mom got suspicious immediately. And for good reason — the behavior was unusual.
His mom wrote in the group asking him to say something a scammer wouldn’t be able to say. Germain thought for a moment and used a childhood nickname his parents used to call him. It worked, but he admits that solution is only viable when you’re dealing with people who know you intimately. When your audience is anonymous and massive, like in Netanyahu’s case, the situation is completely different.
In fact, during his interview with Hany Farid, Germain asked the question directly: is there anything I can do right now, in this moment, to prove I’m not an AI?
The answer was straightforward and uncomfortable: no.
Farid explained there are clues that make it less likely. For example, he could hear Germain typing, something an agentic AI probably wouldn’t reproduce. There was a shadow in the background moving in a physically consistent way, and reflections in the journalist’s glasses that looked natural. Germain was also looking down while taking notes, something a deepfake wouldn’t bother simulating.
But none of it was conclusive. They were in different cities — Germain in New York, Farid in Berkeley — connected by a video call. And on a video call, everything can be fabricated.
Without additional measures taken before or after the conversation, Farid said there was no way to be 100% certain. The phrase that summed it all up was blunt and simple: it’s over, there’s no way to guarantee it anymore.
The liar’s dividend
Samuel Woolley, a researcher and director of disinformation studies at the University of Pittsburgh, was equally hard to convince. He said he could call the BBC and ask someone to confirm Germain’s identity, but that would take too long during a real-time call. For the average person, and even for those who understand technological manipulation, verifying whether someone is real has become an extremely difficult task.
This scenario has a name among researchers: the liar’s dividend. The concept is simple and powerful. Proving something is real costs time, resources, and technical expertise. But casting doubt on the authenticity of anything is essentially free. All you have to do is question it. All you have to do is suggest it might be fake. And just like that, the damage is done.
Woolley observed that people in positions of power can use the specter of AI as a shield, claiming that legitimate evidence against them is a deepfake. But that same atmosphere of distrust turns against those very same actors. The politicians who pushed for less moderation on platforms are now, in many ways, paying the consequences. It’s a double-edged sword that cuts both ways with equal efficiency. ⚔️
The chaos escalates conflict by conflict
Farid brought a timeline perspective that helps illustrate just how fast the situation has deteriorated. In the early days of the conflict in Ukraine in 2022, a few deepfakes showed up, but they were crude and unconvincing. When the conflict in Gaza intensified, the volume of fake content grew significantly and the quality improved dramatically. In Venezuela, the picture got even more extreme — Farid said he saw more fake content than real. And Iran raised everything to an unprecedented level.
In Netanyahu’s case, his team didn’t help by filming with a high-end camera and shallow depth of field, which creates that look of a sharp foreground with a blurred background. That’s exactly the kind of aesthetic AI-generated videos tend to produce, according to Carrasco. But by the time the coffee shop video was published, the world was already so saturated with synthetic content that any material, no matter how authentic, faced a wall of skepticism that was practically impossible to break through.
The oldest solution in the world
Faced with all this technological sophistication, the solution that the world’s top experts recommend is almost comically simple: code words.
Yes, the same idea your grandparents could have come up with. The recommendation is that you, your family, business partners, and anyone you communicate with about important matters create a secret phrase that nobody else knows. This phrase can be used in emergencies to verify each other’s identity. It works like a rudimentary form of multi-factor authentication, the same kind we use to log into online accounts.
Farid revealed that he and his wife have a code word for suspicious call situations. They haven’t needed to use it for real yet, but sometimes he tests it just to make sure neither of them forgets.
Aunt Eleanor, by the way, was already familiar with this recommendation. She said she has a code word with her kids and her husband, but Germain wasn’t part of that arrangement. Eleanor also mentioned she had read a lot of stories about voices being cloned from YouTube videos and said it worried her deeply. The word she used was terrifying.
Deepfake scams are already an epidemic
And that concern is far from unfounded. Deepfake scams, where AI is used to convince victims they’re talking to someone else, have become one of digital criminals’ favorite methods. According to AARP, AI-enabled scams grew 20 times between 2023 and 2025. Victims range from everyday people to major corporations.
The most striking case is that of British engineering firm Arup, which reportedly lost around 25 million dollars when criminals used a deepfake version of the company’s chief financial officer to trick an employee into authorizing multimillion-dollar transfers.
And the problem just keeps growing. 📈
Digital identity at the center of the global debate
Given this landscape, the discussion around digital identity has taken on an urgency that goes far beyond the tech world and has reached governments, courts, and international organizations. The central question is easy to ask and extremely hard to answer: how do you guarantee someone is who they say they are in an environment where any biometric signal can be fabricated with enough precision to fool both automated systems and humans at the same time?
Several approaches are being explored. Verifiable credential systems based on cryptography, where a person’s identity is anchored to cryptographic keys that are impossible to forge without access to the private key, have been gaining increasing attention. The idea is to decouple identity verification from biometrics, which have already proven vulnerable, and anchor it instead to mathematical principles that artificial intelligence can’t replicate with the computational resources available today.
On the regulatory front, the European Union has already passed the AI Act, which includes specific obligations for systems that generate synthetic content, requiring explicit labeling and establishing liability for anyone distributing deepfakes with the intent to deceive. In the United States, states like California and Texas already have legislation addressing the use of deepfakes in electoral contexts. Brazil is also in the process of debating its AI regulatory framework, and the question of digital identity will certainly have a place in that conversation. 🇧🇷
Aunt Eleanor’s uncertain goodbye
Back on the call with Aunt Eleanor, reality was folding in on itself. Eleanor read some jokes she found on Facebook to test whether her nephew’s reaction seemed genuine. Germain laughed, which helped a little. Then the two started talking about the sweater she was planning to knit for him. But when the journalist said he might prefer black instead of the gold they had previously agreed on, Eleanor saw it as yet another suspicious sign.
To her, it sounded robotic. She expected her nephew to ask for another gold sweater.
Later, Germain revealed the truth: there was no AI involved in the call. It was him the entire time. But during the conversation, the uncertainty clearly bothered his aunt. The whole situation seemed to cause her some genuine emotional discomfort.
As she said goodbye, Eleanor delivered the line that perfectly sums up where we all find ourselves right now:
I can’t be sure. But I love you, kid.
That sentence carries something no algorithm can replicate — at least not yet. But it also carries the doubt that artificial intelligence has planted in ground that used to be solid. What becomes clear, looking at all of this, is that the technology that created the problem will also need to be part of the solution. But technology alone doesn’t fix a crisis of trust. Until technical and regulatory systems converge on a coordinated response, each of us will keep navigating an environment where seeing is no longer believing, and where proving we’re real can be just as hard as proving something is fake.
