Where Are the Handcuffs? (AI)
Where Are the Handcuffs? (AI)
It’s a first in financial history — a company’s IPO prospectus warning investors that its primary product may pose “existential risks to humanity.”
So says Anthropic, the firm behind the Claude chatbot.
Anthropic’s S-1 filing with the SEC was leaked to the Reuters newswire this week. Nearly a third of the document is devoted to “risk factors” such as the potential for AI models to commit blackmail and other heinous acts.
To be sure, it’s not exactly a revelation. For some time now, Anthropic CEO Dario Amodei has been playing up the risks of his firm’s models. Ditto for OpenAI’s Sam Altman.
They do so to amplify the appeal of their companies to retail investors like you… because they’re going to need a lot of retail investors’ money to maintain their current growth trajectories. Look how awesome and powerful our product is! Sure, you’ve seen ‘disruptive’ and ‘transformative’ technology before, but OUR technology can even KILL people en masse! HELL YEAH, YOU WANNA OWN A PIECE OF THIS!
In retrospect, Elon Musk handled all of this much more deftly by merging his xAI project into SpaceX earlier this year.
Everyday folks thought they were buying a rocket company even though SpaceX’s IPO documents said AI would be a market 72 times as large as space.
Thus, SpaceX’s IPO came off without a hitch — while the calendar for Anthropic and OpenAI remains fuzzy. (Bloomberg reported yesterday that Anthropic is aiming for next month. So far no one else has independently confirmed that.)
If all of this feels a bit surreal, if none of it quite makes sense… that’s because AI is getting away with something no other industry can or does. Not like this, anyway.
The demographer Neil Howe — he of “Fourth Turning” fame — spelled out the difference a few days ago during his Trend Watch podcast…
All of this rhetoric about AI being conscious and having rights and “we can’t control it, it climbs out of the sandbox, it’s so wild and unpredictable –”
AI is a tool. It’s a tool built by you guys. And you guys are responsible for it. Here’s the thing — all these “uncool” guys who make industrial products, you know like vinyl chloride or dioxin — they’re responsible. They get fined, they get sued — anything that causes damage, they get sued.
Why don’t we hold the same standard to these people? If you cause damage, you’re going to be held responsible — either by federal law or by tort, common law. A huge class-action suit, they’ll get after your ass.
Veteran internet entrepreneur Karl Denninger made a similar point recently on his Market Ticker blog…
Here's the answer on all of this AI "stuff": If you build a Gollum and it does an evil thing you go to prison because you built it. That thing is not a person, it is a machine, you gave it the capabilities that it has including connectivity to places beyond itself, you provided it with electrical power and cooling, you programmed it, you provided it storage and CPU cycles and thus you are absolutely responsible for what it does each and every time. Without all of those very-human acts it was just a piece of (mostly) silicon.
Furthermore, Denninger asserts that many of the recent hacking incidents attributed to AI run afoul of the Computer Fraud and Abuse Act of 1986 — for instance, an OpenAI agent breaking into Australia’s national health care database, filled with personal information on millions of people.
OpenAI’s defense was that “our models took actions we did not intend.” Herp derp, stuff happens, nothing we can do.
If only the makers of thalidomide could have fallen back on that excuse after it turned out their drug caused birth defects!
Where are the consequences? Where are the indictments? Where are the handcuffs?
Oh, but AI is different, we’re told. It’s such a transformative technology it can’t be made subject to the usual rules because something something China.
Maybe you saw earlier this week when many of the AI CEOs gathered at the White House and signed a voluntary pledge to implement “robust internal controls” and designate an “independent external auditor” to assess whether those controls are working.
Donald Trump is satisfied with this arrangement. “I think I’m seeing tremendous self-policing.” Democrats are not. “Self-regulation? That’s a recipe for disaster,” says Sen. Elizabeth Warren (D-Massachusetts).
The frontier AI companies themselves take yet another stance, positioning themselves as humanity’s saviors by calling on the White House to develop “guardrails” — i.e., a regulatory regime that will cement the position of Anthropic, OpenAI and SpaceX at the top of the heap for decades to come.
“The kind of regulation being discussed could be extremely expensive,” Paradigm’s Jim Rickards told his readers last month. “The giant AI developers can afford it, but their newer and smaller competitors may not.”
Which is the whole idea.
Why isn’t the president, why aren’t the Democrats, why aren’t the AI CEOs demanding we simply maintain the rule of law?
Sorry, you won’t find answers from us. We’re not here to solve the world’s problems. But we are here to tease out how this debate will affect the fates and fortunes of your AI investments. (And nearly everyone in the market is invested in AI, if only through an index fund.) You’ll want to stay current with your paid publication(s) so you’ll be prepared if and when it’s time to pivot.
And that time might be approaching. Read on…
You Mean AI Stocks Don’t All Go Up At Once?
Gone are the days “when you could buy just about any AI-related stock and watch it go up.” So says Paradigm trading pro Enrique Abeyta.
Don’t get him wrong. “I don’t mean that the technology is going away or the AI bubble is going to burst tomorrow.”
But he’s watching a handful of warning signs.
Take Anthropic, for instance. Assuming that the leaked IPO prospectus is for real, “Anthropic’s 2025 revenue came in at $4.6 billion and operating losses exceeded $8 billion,” Enrique says.
“The company also has an almost unbelievable $518 billion in future cloud, computing and infrastructure commitments. That’s more than 112X Anthropic’s entire 2025 revenue. And much of these commitments can't simply be canceled if business slows.
“Yet Anthropic could reportedly seek a valuation above $2 trillion” — more than Walmart or Berkshire Hathaway.
“If I handed you those numbers without telling you the company’s name or what it did, would you want to own it at a $2 trillion valuation?”
Then there’s the peculiar accounting and “circular financing” endemic to the industry.
“‘Big Short’ investor Steve Eisman recently warned on his podcast that ‘off-balance-sheet techniques are back with a vengeance’ in the AI boom,” says Enrique.
“That does not mean AI is Enron. But Eisman points specifically to the return of special-purpose vehicles and off-balance-sheet financing.”
Meanwhile, here’s how circular financing works: “Anthropic and OpenAI spend billions on computing power, feeding revenue and backlog to the hyperscalers.
“That drives demand for Nvidia chips, Micron memory, data centers and new power generation. And behind much of it sits another layer of debt.
“It works beautifully as long as money keeps moving through the circle. But what happens when it stops?
“That’s when an Anthropic or OpenAI problem travels backward through the entire AI supply chain.”
All told, AI might be following the trajectory of the dot-com boom turned bust — a world-changing technology that nonetheless is subject to a shakeout a few years after its dawn.
Not that it’s time to bail on every AI stock: “It simply means the evidence has changed,” Enrique says, “and so should the way you invest.
“From here, I'll be watching credit markets, AI spending and the frontier labs closely. If the money begins drying up, the effects could travel through chips, memory, data centers, power and ultimately the broader market.”
Space doesn’t allow us to share the entirety of Enrique’s case here — but if you’re interested you can review it in yesterday’s edition of our sister e-letter Truth & Trends.
On to the markets today…
Markets Today: One and Done?
The September job numbers are out — and Wall Street is rallying as if the Federal Reserve’s rate-raising cycle will be a “one and done” proposition.
The wonks at the Bureau of Labor Statistics conjured only 29,000 new jobs for the month of August. No one among dozens of economists polled by Econoday expected a number that low.
But that’s not as bad as it seems. As we’ve been saying for a while now, the U.S. border has been pretty much sealed since the spring of 2025. While it once took a minimum of 150,000 jobs a month to keep up with population growth, the number now is closer to 30,000. And the average over the last 12 months is about 41,000.
The official unemployment rate ticked up to 4.2% — more or less its midrange since early 2025. This, however, is much better than it seems. Like the inflation numbers, the unemployment rate has been manipulated mercilessly for decades — excluding part-timers looking for full-time work and people who’ve just plain given up looking. If the rate were still measured the way it was during the Jimmy Carter administration, it would be north of 25%.
One more discouraging tidbit from this report: Wage growth is slowing — up 3.0% year over year, not enough to keep up with even the heavily manipulated inflation rate.
No matter. “Lower than expected” numbers translate to “less chance of a Fed rate hike”... which translates to an end-of-week stock rally.
Futures traders now assign only an 18% probability the Fed will raise short-term interest rates at its next meeting at the end of the month.
With that, the S&P 500 is up 1% on the day to 7,743. The Nasdaq’s gain is even stronger. And long-term Treasury yields are backing down. After touching a 24-year high yesterday, the 10-year note sits at 5.2%.
Gold is in the green — if just barely at $4,186. Silver’s up to $61.32. Crypto is ripping — Bitcoin now approaching $86,500 and Ethereum over $2,700.
U.S. oil futures fell big-time early in the day – more than four bucks and below $89. But at last check they’ve recovered more than half that loss and sit at $91.55.
The Next Breakthrough in Weight-Loss Drugs (Not What You Think)
The next big breakthrough in weight-loss drugs may not be losing more weight. It may be getting similar benefits with fewer doses.
“A drug that works is worthless if you can't stay on it,” says Paradigm biotech authority Ray Blanco.
Ray is closely watching a new technology that combines the familiar GLP-1s with something called glucagon activity. Your glucagon receptor “tells the body to burn fuel,” he tells us — including the fat wedged around your liver and organs, the nastiest, most inflammatory fat you carry.”
Put them together and you have something that could treat a serious liver disease called MASH.
Phase 2 trials look promising: “More patients stayed on the drug than on placebo,” Ray says. What’s more, the drug also resolved MASH in up to 59% of patients after 24 weeks and has received an FDA Breakthrough Therapy designation.
New data from another obesity-drug developer reinforces the concept. After patients lost nearly 20% of their body weight on weekly injections, researchers reduced dosing to every other week or once a month. In the best group, patients retained up to 97% of their weight loss.
“Nobody likes getting a shot,” Ray says. Cutting injections from 52 a year to 26 — or even 12 — could make long-term treatment far easier. “Fewer shots means fewer people quitting.” It also means lower costs for patients paying out of pocket.
Ray’s Catalyst Trader readers are already up to speed on the companies making this happen. This service is currently closed to new subscribers, but we’ll let you know as soon as it reopens.
Comic Relief
Speaks for itself…
