It was a Tuesday morning when I got the call. The CEO of a promising startup in my portfolio—a company I’d backed with my own money—was on the other end of the line, and he sounded panicked. They had been hit. Not by a traditional data breach, but by something new, something insidious. An attacker had used an AI-generated voice clone of the CEO to authorize a six-figure wire transfer. The clone wasn’t perfect, but it was good enough to fool a junior employee in finance. They caught it just in time, but it was a brutal wake-up call. For them, and for me.
Over the past two years, I've reviewed the post-mortems of over 100 AI-related security incidents, and the story is always the same: we are building technological skyscrapers on foundations of sand. As an entrepreneur who has been through two exits and an investor with over 200 bets in tech—including foundational companies like Anthropic, OpenAI, and Scale AI—I’ve seen my share of hype cycles. But the explosion in AI capabilities isn’t just hype. It’s a fundamental shift in how we build, and it’s creating a security debt that will be astronomically expensive to pay back.
Everyone is so mesmerized by what AI can do that they’re not thinking about what it can be forced to do. And that’s where the next trillion-dollar opportunity lies.
The New Wave of Threats is Already Here
Forget the old playbook of cybersecurity. The tools that protect us from traditional malware and phishing are like bringing a knife to a gunfight when it comes to AI-native threats. The attack surface is no longer just our networks and servers; it’s the logic of the models themselves. I’ve seen the risks firsthand, and a few key patterns keep emerging.
1. Phishing on Steroids
The voice clone incident was just the beginning. We’re now seeing AI-powered phishing campaigns that are terrifyingly personal and effective. These aren’t your typical "click here to reset your password" emails. Imagine an email that perfectly mimics your boss’s writing style, references a private conversation you had last week, and asks you to review a document that contains a subtle, AI-generated payload. It’s a level of social engineering that humans alone could never scale. These attacks don’t just trick people; they manipulate them by hijacking trust.
2. The Rise of "Zero-Day AI" Exploits
In traditional security, a "zero-day" is a vulnerability that’s unknown to the good guys. In the world of AI, the concept is even scarier. A zero-day AI exploit isn’t about finding a bug in the code, but about finding a flaw in the model’s reasoning. It’s about crafting an input that seems harmless to a human but causes the AI to behave in a completely unexpected and dangerous way.
Think of it like a visual illusion for a machine. One of the most common examples is a "prompt injection" attack. An attacker can hide a malicious instruction inside a seemingly benign piece of text. When a large language model processes that text, it executes the hidden command—leaking sensitive data, generating harmful content, or bypassing its own safety filters. Traditional firewalls can’t see this. They’re not built to understand intent.
3. Data Poisoning: Corrupting the Source of Truth
Every AI model is only as good as the data it’s trained on. So what happens when an attacker intentionally poisons that data? I saw one case where a competitor subtly fed a retail company’s pricing AI thousands of fake data points over several months. The goal was to slowly teach the model that a specific product category should be priced far below its actual cost. When the company ran its next big promotion, the AI priced the items at a 90% loss, and they lost millions before they figured out what had happened. The attack wasn’t a frontal assault; it was a slow, deliberate corruption of the AI’s "brain."
Why I’m Betting Big on AI Security
After my first exit with MovieLaLa, I learned to spot patterns. After my second with RemoteTeam, I learned that the biggest opportunities are often found in the problems everyone else is ignoring. The current gold rush in AI is focused on building bigger, faster models and flashier applications. Very few are building the picks and shovels needed to keep the gold miners safe.
This is not a niche for the existing cybersecurity giants to simply add a new feature. Defending against these threats requires a fundamentally new approach. It requires a deep, almost intuitive understanding of how models think, learn, and fail. It’s a field where a small team of brilliant researchers can run circles around a massive incumbent.
As an investor, I’m actively looking for the startups that get this. I’m looking for founders who have spent time in the trenches of both AI research and cybersecurity. The teams that will win are not the ones bolting AI onto an old security product. They are the ones building security from the ground up, designed for a world where your AI can be turned against you.
We need tools that can detect AI-generated phishing in real-time. We need firewalls for models that can spot and block prompt injection attacks. We need systems that can continuously validate the integrity of training data to prevent poisoning.
This is more than just a market opportunity; it’s a necessity. The next generation of technology is being built on AI, and without a robust security layer, we are setting ourselves up for a systemic failure. The entrepreneurs who solve this won’t just build successful companies; they will be safeguarding the future of technology itself. If you’re a builder in this space, I want to hear from you. This is the moment to get ahead of the curve and build something that matters. It’s your chance to become part of the top 1%.
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.
What's the most common pushback you get on this?
People often push back by citing exceptions or edge cases. And they're usually right that exceptions exist. But building a strategy around exceptions rather than patterns is a losing game for most founders.
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.
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.