I almost lost a portfolio company to a phone call.
It wasn't a bad business deal, or a market crash, or some other "act of God". It was a 30-second phone call that almost convinced a junior employee to wire $250,000 to a fraudulent account. The voice on the other end? A perfect, real-time clone of the CEO's voice. We caught it in time, but it was a wake-up call for me. As an investor in over 200 companies, including some of the biggest names in AI like Anthropic, OpenAI, and Scale AI, I've had a front-row seat to the incredible advancements in this field. But I've also seen the dark side. And let me tell you, the threat that keeps me up at night isn't deepfakes. It's real-time voice cloning.
The Unseen Threat
For the past few years, all the headlines have been about deepfake videos. And for good reason. The ability to create realistic videos of people saying and doing things they never did is terrifying. But here's the thing: creating a convincing deepfake video is still hard. It takes time, data, and a lot of computing power. And most importantly, it's not real-time. There's a delay, a rendering process.
Real-time voice cloning, on the other hand, is here. It's happening right now. And it's scarily easy to do. With just a few seconds of someone's voice, you can create a clone that is indistinguishable from the real thing. You can then use that clone to say anything you want, in real-time, with the same intonation, emotion, and accent as the original speaker.
Think about the implications of that for a second. You could have a "conversation" with someone who isn't there. You could authorize a financial transaction, give a sensitive order, or even confess to a crime, all without ever opening your mouth. The barrier to entry is collapsing. Gone are the days when you needed a sophisticated lab and a team of audio engineers. Now, all you need is a laptop, an internet connection, and a few lines of code. There are open-source projects on GitHub that can get you started in minutes. For a few dollars, you can even rent cloud computing power to create highly convincing clones. It's a far cry from the multi-million dollar setups that were required just a few years ago.
Beyond the "Grandma Scam"
The media loves to talk about the "grandma scam," where a fraudster calls an elderly person, pretends to be their grandchild, and asks for money. And yes, that's a real and growing problem. But it's just the tip of the iceberg. The real threat is to businesses, financial institutions, and even national security.
I've seen this firsthand. In the case of my portfolio company, the fraudsters had done their homework. They knew the CEO's name, the employee's name, and the company's bank details. They even knew that the CEO was traveling and would be hard to reach. The only thing that saved us was the employee's gut feeling that something was off. The CEO's "voice" sounded a little too perfect, a little too smooth. But what happens when the technology gets even better? When it can perfectly replicate the ums, the ahs, the pauses, and the imperfections of human speech? What happens when it can do it in any language, with any accent?
We're not talking about a distant future here. We're talking about now. I was recently talking to a founder of a cybersecurity startup I invested in. They ran a test. They took a 30-second clip of my voice from a podcast I did and used it to clone my voice. They then called me and played the clone. It was so good that for a few seconds, I thought I was talking to myself. It was a chilling experience.
The New Frontier of Fraud
We're already seeing the next wave of voice cloning scams. They're more sophisticated, more targeted, and much harder to detect. Here are a few of the things I'm seeing:
- CEO Fraud 2.0: This is the evolution of the scam that almost hit my portfolio company. The fraudsters are now using real-time voice cloning to impersonate CEOs and other executives in live phone calls and video conferences. They're using this to authorize fraudulent wire transfers, gain access to sensitive data, and even manipulate stock prices. A recent report from a cybersecurity firm found that CEO fraud attacks using voice cloning have increased by over 350% in the last year alone. The average loss from these attacks is now over $500,000.
- Vishing on Steroids: Vishing, or voice phishing, has been around for a while. But with real-time voice cloning, it's becoming much more effective. Fraudsters can now impersonate anyone, from a bank teller to a government agent, with terrifying accuracy. They're using this to steal personal information, credit card numbers, and even social security numbers. I've heard stories of people losing their life savings to these scams. They get a call from their "bank," the voice on the other end sounds exactly like the friendly teller they've been talking to for years, and they hand over their information without a second thought.
- Political Manipulation: This is the one that really scares me. Imagine a world where you can't trust anything you hear. Where a political candidate can be made to say anything, in their own voice. Where a world leader can be impersonated to start a war. It sounds like science fiction, but it's not. The technology is already here. We've already seen examples of this in smaller countries. In one case, a fake audio clip of a political candidate admitting to corruption was circulated on social media just days before an election. The clip went viral, and the candidate lost the election. It was later proven to be a fake, but by then, the damage was done.
The Arms Race of AI
This is the new reality we live in. It's an arms race. On one side, you have the black hats, the criminals and the state-sponsored actors who are using this technology for nefarious purposes. On the other side, you have the white hats, the cybersecurity experts and the AI researchers who are trying to build defenses against it.
I'm an optimist. I believe that in the long run, the white hats will win. But it's not going to be easy. The black hats have a significant advantage. They don't have to play by the rules. They don't have to worry about ethics or regulations. They can move fast and break things. The white hats, on the other hand, have to be more careful. They have to make sure that the defenses they build don't have unintended consequences. They have to worry about things like bias and fairness.
This is why I'm investing in companies that are working on both sides of the problem. I'm investing in companies that are building better voice cloning technology, because I believe that this technology has the potential to do a lot of good in the world. But I'm also investing in companies that are building better defenses against it. We need both. We need to push the technology forward, but we also need to make sure that we have the tools to protect ourselves from its misuse.
So, What Do We Do?
I'm not going to lie, the problem is a big one. And there's no easy solution. But there are things we can do to protect ourselves.
First, we need to be more skeptical. We need to question everything we hear, especially when it comes to financial transactions or sensitive information. If you get a call from your "CEO" asking you to wire money, hang up and call them back on their known number. If you get a call from your "bank" asking for your password, don't give it to them. It's simple advice, but it's the most effective defense we have right now. We need to build a culture of security awareness, where everyone in an organization feels empowered to question a request, no matter who it's coming from.
Second, we need to invest in new security technologies. There are companies working on developing "voice-printing" technology that can distinguish between a real voice and a clone. There are also companies working on developing new authentication methods that don't rely on voice alone. For example, some companies are experimenting with multi-factor authentication that combines voice with other biometrics, like facial recognition or fingerprint scanning. Others are working on developing AI-powered systems that can analyze the context of a conversation to detect anomalies. These are all promising developments, but they're still in their early stages. We need to accelerate the research and development in this area.
Third, we need to educate ourselves and our employees. We need to make sure that everyone in our organization understands the threat of real-time voice cloning and knows how to spot a scam. This is not just a problem for the IT department. It's a problem for everyone. We need to run regular training sessions, send out security alerts, and create a culture where people feel comfortable reporting suspicious activity. I've implemented a simple rule in my own companies: any financial transaction over a certain amount requires a two-person approval, and one of those approvals has to be in person or via a video call where the person's identity can be verified.
I know it sounds like a lot. And it is. But the threat is real, and it's only going to get worse. We can't afford to ignore it. The future of our businesses, our finances, and even our democracy may depend on it. This is not a problem that we can solve with technology alone. It's a human problem. It's about trust, and how we verify it in a world where our senses can be so easily deceived. The conversation is just beginning, and we all need to be a part of it.
Frequently Asked Questions
What experience informs this perspective?
This perspective comes from over a decade of building companies in Silicon Valley, two successful exits (RemoteTeam to Gusto, MovieLaLa to Gfycat), and investing in 200+ startups including Anthropic, OpenAI, and Scale AI. I write about what I've lived.
Do all experts agree with this view?
No, and that's fine. The best ideas in business are often contrarian. I share my perspective based on my experience and data, but I encourage you to seek out opposing viewpoints and form your own conclusions.
How can I apply this thinking to my own situation?
Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.