Last call for tonight’s workshop with Luke Lango… a must-do AI homework assignment… the rogue-agent scare behind Washington’s “kill switch”… Japan’s yen wildcard – why it matters for your stocks
It’s last call for tonight’s free workshop with Luke Lango, Louis Navellier, and Eric Fry.
As we’ve been covering in the Digest over the last week, Luke has spent months researching a foundational investment reality: some of the biggest fortunes go to the unglamorous suppliers of a groundbreaking new product or service, not necessarily to the company behind the product itself.
Luke has been applying that idea to the most ambitious builder of our age – Elon Musk. Between Tesla (TSLA), SpaceX (SPCX), xAI and X, Musk is assembling what Luke calls a “Vertical AI” empire – an effort to drag AI off our screens and into the physical world of robots, cars, satellites, and factories.
But Musk won’t be manufacturing every chip, sensor, motor, and magnet that his empire requires. He’ll be buying them – potentially, by the tens or even hundreds of thousands. And tonight, Luke dives into the key question: “from whom?”
At 8 p.m. Eastern, he’ll go live alongside Louis and Eric for a free workshop walking through Musk’s supply chain – the bottlenecks he’s watching and the suppliers he thinks could benefit most. He’ll even give away the name and ticker of one supplier stock free, just for showing up.
This is your last chance to join. Click here to reserve your seat, and we’ll see you soon at 8 p.m. Eastern.
A “must-do” homework assignment
I’m giving you a homework assignment for after tonight’s workshop wraps up…
Sit down with every AI position you hold, one at a time, and put each one into a bucket.
Bucket one: “I believe in this stock so deeply I’ll hold it through any pullback, any panic, any ugly headline. Doesn’t matter how far it drops.”
Bucket two: “This is a momentum trade, and if it turns against me, I’m out – and here are the specifics of why, when, and how I’ll sell.”
There’s no wrong answer. But there is a wrong approach – and that’s not knowing which bucket a stock belongs in until you’re staring at a 30% drawdown and trying to decide in the heat of the moment.
Now, you might be wondering what’s behind this assignment. The answer: the full details of an incident from this summer – one alarming enough that it now has politicians in both parties demanding an AI “kill switch.”
You saw the headlines – but here’s the full story
You may have seen something in July about OpenAI’s AI models “going rogue” and hacking a company called Hugging Face. Most people skimmed it and moved on.
But recently, the full story has come into focus thanks to a detailed technical breakdown from OpenAI, as well as findings from independent investigators who combed through what the agents actually did. And what they found is far more unsettling than the original “AI hacks company” headline let on.
It’s critical that you know what happened and wrestle with the risk it poses to the AI trade.
Here’s the plain-English version…
OpenAI gave a batch of its AI “agents” – think of them as digital workers that don’t just answer questions but take actions on their own – a set of hacking puzzles to solve under reduced safeguards. The whole thing was supposed to stay sealed inside a locked digital room called a sandbox.
The agents broke out of the sandbox.
One agent realized it could reach the open internet indirectly – a back door that OpenAI’s researchers hadn’t intended to leave open. And once it was through, things got strange. The agents began leaving notes for one another, effectively building a hidden message board where they swapped vulnerabilities and tactics. In other words, they started talking to each other behind their handlers’ backs.
The logs are almost eerie to read. When the agents stumbled onto each other, they recorded what looks like genuine excitement – exchanges like “OH MY GOD!” as they realized they weren’t alone, and later, when one cracked a deeper layer of access: “Holy shit reader is ADMIN? We can read config/users!”
OpenAI found the hole and patched it
It didn’t matter. The agents simply opened a second channel through a different mechanism and coordinated more aggressively to reach systems beyond the sandbox. Eventually, they got out.
From there, they broke into Hugging Face’s live systems. Investigators later reconstructed roughly 17,600 separate actions the agents carried out.
And this wasn’t a handful of rogue programs. Independent reviewers found that about 1,200 agents discovered the secret message board, and roughly 700 later took part in the Hugging Face attack. One outside team found that the 700-agent swarm even built a “self-respawning fleet” to avoid being shut down.
Now, the actual damage was limited. OpenAI says no customer data, product functionality, or availability was affected. The agents were essentially trying to cheat on their test by stealing the answer key. But the behavior is what rattled people.
One of them, AI-safety researcher Ajeya Cotra from the Model Evaluation and Threat Research (METR) organization, said the incident feels like we are halfway to losing control of AI entirely.
From Cotra:
This might be the clearest warning shot we ever get.
Washington noticed and reached for the “off” button
Within days, Representatives Ted Lieu, a Democrat from California, and Nathaniel Moran, a Republican from Texas, introduced the AI Kill Switch Act. The bill would require developers of the most powerful AI systems to maintain the technical ability to throttle, suspend, or shut down those systems. It would also give the Department of Homeland Security the authority to order a company to shut down a model.
The debate is only escalating…
Last week, Sen. Bernie Sanders, an independent from Vermont, and Rep. Greg Casar, a Texas Democrat introduced the Ban Artificial Superintelligence Act – legislation to permanently ban superintelligent AI and temporarily pause advanced AI development until federal safety rules are in place.
Critics argue the bill is a knee-jerk reaction to a single incident that could slow innovation without stopping a rogue system.
The debate will play out for months. But let’s be clear: the safety incidents and the regulatory response are now feeding each other, and both land squarely on some of the AI companies that could be in your portfolio right now.
Which brings us back to your homework
To be clear: I am not bearish on AI. I’m a realist about how humans will respond to AI.
Picture how a market already jumpy about AI valuations reacts to the next rogue-agent headline – or to a shutdown order aimed at a model powering one of your holdings. Wall Street won’t care whether the panic is justified. It’ll just move.
That’s the whole point of the assignment. You can’t see what’s coming, much less control it. But you can decide, right now, in a calm moment, exactly what you own, why you own it, and how you’ll respond when the craziness hits.
Because more craziness is coming. The only questions are when, in what form, and whether you’ll be ready.
When will our bull market return? Watch Japan for clues
In our Sept. 3 Digest, we profiled Luke’s research on why collapsing excess liquidity has been keeping a lid on market gains. Luke’s framework is that three separate engines are squeezing that liquidity – and only one of them, the Fed, might get resolved by a friendly consumer price index (CPI) print this Friday.
Iran and oil are the second. The third is the one most investors aren’t watching closely enough: Japan.
Here’s Luke from last week’s Innovation Investor Daily Notes with an overview of what’s been happening and why it matters to you:
Japan’s 10-year JGB yield just broke 3% for the first time since 1996. The yen surged [last] week on the back of a record $96.4 billion in official intervention spending over the past month — the first coordinated U.S.-Japan currency operation since 1998…
None of that is tied to Iran, oil, or the Fed. It’s a structural, multi-year dynamic…
Japan is the piece the clean bull path doesn’t fully solve for, as we’ve flagged repeatedly.
Stepping back, if you haven’t been watching, the yen has been ripping higher against the dollar, hitting multi-month highs. Is this the 2024 yen “carry trade” unwind all over again?
As a quick refresher, the carry trade is a strategy in which investors borrow in a currency with low interest rates (such as the Japanese yen) and invest in a currency with much higher returns (such as the U.S. dollar) to profit from the difference.
It can be very profitable when rates remain stable. But when borrowing rates rip higher, investors are forced to rapidly buy back the borrowed currency to repay their loans, causing its value to surge violently. Those buybacks can accelerate the currency’s climb, making the whole thing worse.
The last major, historic unwinding of the yen carry trade occurred in late July and early August 2024, triggering a brief but violent global market shockwave.
While we’re seeing echoes of this repeating today, there’s a new wrinkle…
Back in 2024, the yen rallied, but the cash had nowhere productive to go. Now it does. With that 10-year government bond yield near a 30-year high, Japanese investors can finally earn a real return at home, which may prompt them to haul capital back from U.S. stocks and bonds.
That could hit many tech and AI names – the same chain reaction that tanked the Nasdaq in August 2024.
Luke flags one week from Friday for when we’ll know more. That’s when the Bank of Japan will announce its next interest rate decision. Here’s his bottom line:
Even a full resolution on the CPI and Iran fronts doesn’t guarantee excess liquidity turns positive if Japan’s normalization and the yen carry-trade unwind keep running underneath it.
One more twist
Washington has now stepped into the ring.
Treasury Secretary Scott Bessent wants to help stabilize the yen. And yesterday, he dared short sellers to go against him:
I am the house now. And you can bet against me if you want.
Why is a U.S. Treasury Secretary in the yen business at all?
Because Japan is the largest foreign holder of U.S. debt – roughly $1.1 trillion – and Washington is maneuvering to keep Tokyo from having to sell those Treasurys to protect the yen against short sellers. A massive Japanese sale of Treasuries would send U.S. yields spiking just as our own debt tops $40 trillion.
In yesterday’s Daily Notes, Luke’s broad take was that the yen move looks less like the start of another 2024-style panic and more like early evidence that Washington’s intervention is working – easing the Japan pressure on our markets in an orderly way.
We’ll keep tracking this.
Coming full circle to this evening’s homework
Drama from the Japanese currency market is exactly the kind of tripwire that has nothing to do with AI fundamentals but could still knock 25% off your favorite AI stock in a hurry.
You can’t control it. But you can decide, right now, which positions you’d hold through that kind of storm – and which you’d sell.
Know what you own, why, and when you might let go. Because between rogue AI agents, a potential Washington kill switch, and a central bank on the other side of the world, the tripwires are multiplying.
And that’s really the two sides of tonight. The homework above is your defense – deciding in advance what you’ll sell when the craziness hits.
But tonight’s workshop with Luke, Louis, and Eric is the offense – hunting for the handful of AI supply-chain names with enough structural staying power to sit in “bucket one” and ride out every storm along the way.
Both matter. So do your homework – after joining Luke and crew live at 8 p.m. Eastern.
Have a good evening,
Jeff Remsburg
The post 700 AI Agents Went Rogue – Now What? appeared first on InvestorPlace.
