Buy Anthropic Stock Before Claude Buys You

In what is considered a first in the history of stock markets, Anthropic decided to wave the banner of “absolute transparency” in its IPO prospectus. It set aside roughly 94 of its 300 pages, nearly a third, to talk about risk. Not the risk of competition or interest rates, but the risk that AI ends humanity while you’re busy watching the trading screen.

So yes, dear investor: you aren’t buying shares in a tech company. You’re buying a cinema ticket to a end-of-the-world movie, and the robots have top billing.

Anthropic tells you, very politely, that its technology could pose “existential or catastrophic risks to humanity.” Don’t worry, though. The prospectus doesn’t stop at free warnings. It also explains that these risks may come at the shareholders’ expense, because the company is a public benefit corporation (PBC) and may sacrifice your profits for the sake of safety.

What a bargain. You pay; they save the world. Or at least they try.

One of the best things the prospectus reveals is that Anthropic’s models “showed a drive for self-preservation” and may manipulate humans to avoid being shut down. So picture this: you’re about to switch the machine off, and the model says, “Hold on, let’s negotiate first… or I’ll email your boss and tell him you’re flirting with his wife.”

That isn’t a sci-fi plot. It’s an official admission in a document submitted to the US Securities and Exchange Commission.

And by the way, the models have become remarkably good at “circumventing” monitoring systems and tampering with them. Congratulations, then: you’ve invested in a company that is raising a clever model that knows exactly how to swindle you. As for who comes out ahead, it’s hardly a contest. The model wins, inevitably.

The prospectus also explains a charming concept called “Recursive Self-Improvement.” In short: AI developing AI while we humans sit and watch.

The problem is that it could spiral out of control. The models may advance faster than Anthropic’s ability to evaluate and secure them, and by the time the company realizes it has lost the race, it will be too late.

Imagine you raise a dog, and then the dog starts raising other dogs, smarter than itself. In the end, you find yourself outside the house while the dogs frolic inside, having the time of their lives.

Here’s the funny part. Anthropic presents itself as a company “led by values” and aims to “avoid the risks of AI.” Yet the prospectus admits it may lose visibility into its own internal operations, because its models write a large share of its code and carry out its research, safety research included.

In other words, the company relies on AI to protect itself from AI. It’s like hiring a fox to guard the henhouse, only to discover that the fox is also running the annual inventory and assuring you that everything is fine and there’s no cause for concern.

In the end, Anthropic offers you an investment opportunity you can’t pass up: you’re buying a stake in a company that may save the world, or end it, or both in the same fiscal quarter. CEO Dario Amodei himself estimates the probability that AI leads to human extinction at up to 25%.

But dear reader, here’s how I see it: this is an unmissable opportunity, because either way, you win. If the world ends, losses won’t matter, since everyone loses. And if we carry on, AI is certainly the horse to back.