A few years ago, the big question about artificial intelligence was simple:
When’s AI going to take my job?
Then came the robots writing code. Then came the AI agents booking meetings, doing research, and eventually, running entire businesses.
But now the question has gotten darker:
How long do we have before AI kills us all?
This may sound like science fiction. But over the past two weeks, the people building this technology have started talking about it seriously…
And that creates a fascinating opportunity for investors like you.
From Jobs to Judgment Day
It all started when Jacob Coxon, a former Anthropic and OpenAI researcher, resigned and posted on X that AI companies were “racing straight to self-improving superintelligence.”
The post went viral. It’s racked up more than 150 million views so far.
Then Evan Hubinger, Anthropic’s Alignment Science Lead, responded with an even more startling assessment. He said he personally believes there’s greater than a 10% chance that AI could kill all humans within the next decade.
His response has gotten nearly 50 million views so far.
Suddenly, AI safety wasn't some obscure concern discussed by researchers. It was front-page news. And then something even stranger happened:
Anthropic CEO Dario Amodei called for the industry to slow down AI development. OpenAI CEO Sam Altman and xAI's Elon Musk backed the idea, while other major AI players quickly joined the push for greater safety and oversight.
So what should an investor like you make of all this?
Three Possibilities
There are, it seems to me, three possibilities.
First, all these people truly believe it. Maybe the people closest to the technology genuinely think we're approaching a point where AI systems could become too powerful to control.
Second, it could be self-interest. Slowing the pace of development could have convenient side effects for the companies involved — including buying them more time to deal with enormous computing costs, product-development challenges, regulatory scrutiny, and upcoming IPOs.
That doesn't prove the safety concerns are manufactured. It simply means the incentives are complicated.
Or, most likely, the truth is somewhere in between. The risks could be real — and the companies could still benefit financially from slowing the race.
But here's the interesting part for investors:
No matter which explanation is correct, we need security systems to control AI.
The Obvious Winners Are Already Expensive
The first group of potential winners here is the big, publicly traded cybersecurity companies:
CrowdStrike, Palo Alto Networks, Okta, Zscaler, SentinelOne, Fortinet and Cloudflare.
The problem is, a lot of investors have already figured this out.
That’s why, on September 17, investment bank Bernstein downgraded Palo Alto Networks, Okta and SentinelOne from Outperform to Market Perform, arguing that cybersecurity valuations had become increasingly stretched after a huge run-up.
That’s why I'm looking somewhere else — startups.
The Real Opportunity: AI Security Startups
From a profit perspective, I believe the far more interesting opportunity is the next generation of companies being built specifically to secure AI.
Take Neo, for example. The company emerged from stealth this summer with $100 million from Andreessen Horowitz and Bessemer Venture Partners. Its technology is designed to give enterprises visibility and control over AI agents and other agentic software.
Then there's Abnormal Security, which uses behavioral AI to protect enterprises from sophisticated cyberattacks, and has expanded into AI governance and AI security. The company has already surpassed $200 million in ARR.
Cyberhaven is tackling another problem: data. Its technology tracks how sensitive data moves through enterprises — including through AI systems and autonomous agents — and helps companies prevent that data from leaking. The company raised $100 million at a $1 billion valuation in 2025.
Protect AI is focused directly on securing machine-learning systems against threats such as prompt injection and data poisoning.
HiddenLayer goes after adversarial attacks, model manipulation, and runtime threats. It just raised another $100 million to expand its AI-security platform.
And Onyx Security is building what it calls a secure AI “control plane” — essentially infrastructure for discovering, monitoring, and governing autonomous AI agents. It launched with $40 million in funding.
Follow the Fear
Here's the lesson for investors:
Whenever a transformative technology creates a new problem, it also creates an enormous new market for solving that problem.
In other words, the bigger AI gets, the bigger the AI-security market becomes.
And if the people building AI are right about the risks, security could become one of the most important software categories of the next decade.
That's why, with my investor hat on, I'm less interested in asking, “Will AI kill us all?”
And more interested in asking, “Who gets paid to ensure it doesn't kill us?”
These are the kinds of early-stage companies we're looking for at Crowdability:
Startups attacking enormous emerging markets — before the opportunity becomes obvious to everyone, and before the valuations get bid up accordingly.
Happy Investing,
