Grandfather Fooled by His Brother's Voice—His Granddaughter Decided to Fix It Herself

One day, a phone call came in for Tarini Puri's grandfather.

On the other end was her grandfather's brother's voice — the familiar accent, the rhythm of his speech, even the intimate terms of address that only old friends use between themselves. Her grandfather had no reason to doubt it. He transferred the money. It wasn't until they actually reached his brother that they discovered he had never made that call.

This isn't science fiction. According to a TechCrunch report from September 2026, this happened in Tarini's own family — and her response wasn't to post an angry rant on social media. She went and built a company.

Why This Is Harder to Prevent Than You Think

You might say: "Just ask them to video call, then."

The problem is that deepfake technology in 2026 goes far beyond voice synthesis — real-time face-swapping in video is also achievable, and it requires less and less training data to pull off. Three to five minutes of recorded conversation is enough to train a voice clone capable of speaking in real time. Has your grandfather ever posted videos on TikTok or YouTube? Appeared in a livestream of a funeral or wedding? All of that audio is collectible.

What makes this worse is that the people being targeted are usually the elderly — not because they're naive, but because their default setting is trust. When you've known a voice for decades, you don't think to question it first. That's precisely the vulnerability that scammers exploit.

What Tarini Did — More Useful Than "Telling People to Be Careful"

She didn't stay in the role of victim. She researched the field of voice authentication, then founded a company focused on helping families and individuals establish a voice identity verification mechanism. In plain terms: when a "suspicious" call comes in, the system can verify whether the caller's voiceprint matches the real person — not just whether it sounds like them.

This isn't an entirely new idea. The financial industry has long used voiceprints for call center authentication (if you've ever called certain banks, you may have heard "your voice is your password"). But bringing that technology down to the household level — making it usable by elderly people who aren't tech-savvy — is the real challenge.

Think of it as a two-layer moat: the first layer is detection (is this voice AI-generated?), and the second is verification (even if the voice is real, is it actually the person you know?). Most solutions on the market today only address one of these layers. What Tarini is targeting is integrating both — with a low enough barrier that anyone can use it.

The Scam Industry Is Upgrading Faster Than Regulation

Here's a reality that I find somewhat discouraging: the technology is moving too fast.

In 2024, global losses from AI voice scams were estimated at over $25 billion — and that figure is accelerating. Regulatory bodies tend to react slowly; by the time a country enacts meaningful legislation, attackers have already moved on to a new set of tools.

The proliferation of AI Agents makes things even more complicated — today's AI can not only generate voices, but automate entire conversation workflows, including gathering intelligence on targets in advance, selecting the optimal moment to strike, and then executing the attack. You're no longer dealing with a single scammer. You're up against an automated fraud pipeline running simultaneously against thousands of targets.

This is precisely why founders like Tarini matter: she isn't waiting for regulation. She's building a product.

What You Can Do (Now, Not Later)

A few things that can actually be put into practice:

• Establish a family code word: Agree on a "real person confirmation phrase" with your elderly relatives. If the caller can't produce it, no money gets transferred — no matter how convincing the voice sounds. It's low-tech, but it works.
• Be alert to manufactured urgency: Nearly every voice scam shares one common feature — artificial time pressure ("I'm in the hospital right now," "You need to transfer the money immediately"). The moment that dynamic appears, hang up and call back using the number you already have saved.
• Don't put your voice samples out carelessly: This isn't about disappearing from the internet. It's about recognizing that your voice is a biometric that can be copied.
• On the tools side: Several browser extensions and mobile apps now offer real-time detection of AI-synthesized voice characteristics during phone calls. They're worth searching for and installing.

As an aside, if you're thinking about how to make communications safer within your organization, the choice of AI meeting note tools also involves questions of who can access conversations and how data is stored — a dimension that often gets overlooked.

One Thought to Take With You

Tarini's story brings to mind a saying: "The best security products usually come from people who've personally stepped on the landmine."

Her grandfather's experience was a real loss — but it forged a founding motivation more resilient than any market opportunity analysis or investor pitch deck. Look into the companies that have made meaningful progress in AI security, and behind many of them you'll find a concrete story that kept the founder up at night.

I'm not romanticizing victimhood. What I'm saying is: if you have elderly relatives, it's worth spending five minutes at dinner tonight talking about this. Not to frighten them — but to let them know that "that voice sounds exactly like him" no longer means it is him.

AI has made many things faster and better. But it has also made a phone call far harder to trust than it used to be.

References

• After a deepfake voice fooled her grandfather, this founder sprang into action — TechCrunch (2026-09-28)