Whitney Webb's Chilling Theory About the Next Financial Crisis, The AI Poisoning Attack That Should Terrify You, Why Japan's Bond Market Should Terrify Everyone | Weekly Recap

What's up, everybody? It's Tom Bilyeu here:

Want my help starting a business? Join me here inside Zero To Founder

Sign up for my AI Masterclass: AI Masterclass

FOLLOW TOM:

Instagram: https://www.instagram.com/tombilyeu/

Tik Tok: https://www.tiktok.com/@tombilyeu?lang=en

Twitter: https://twitter.com/tombilyeu

YouTube: https://www.youtube.com/@TomBilyeu

Tailor Brands: Check out Tailor Brands to get started with your business today: https://bit.ly/TailorBrandsSept

Quince: Free shipping and 365-day returns at https://quince.com/impactpod

ElevenLabs: Book your demo at https://elevenlabs.io/impactpod

Incogni: Take your personal data back with Incogni! Use code IMPACT at the link below and get 60% off an annual plan: https://incogni.com/impact

Ketone IQ: Visit https://ketone.com/IMPACT for 30% OFF your subscription order.

Horizon.ai: Go to https://horizon3.ai/IMPACTTHEORY and request your free NodeZero demo. No commitment required. Results in hours, not weeks.

Butcherbox: Go to https://ButcherBox.com/IMPACT to get $20 off your first box, plus your choice of free ribeye, new york strip, or filet mignon in every box for a year — with free shipping always

Quo: ​​Try for free PLUS get 20% off your first 6 months at https://quo.com/impact

The team reacts to a clip from journalist Whitney Webb, who lays out a striking thought experiment about how a future financial crisis could unfold. Her scenario: a cyberattack on the banking system—whether real or manufactured—could let banks absolve themselves of blame for a crisis, consolidate power into a handful of "systemically important" institutions, and serve as a mechanism to onboard the public onto a new digital-currency paradigm. The chilling version: your money "disappears" in the hack, and you're told you can have the exact same amount back—but only in an approved digital dollar, stablecoin, or CBDC, with new conditions attached. You "voluntarily" accept the new system, but under extreme coercion. The host is careful and explicit throughout that this is Webb's thought experiment, not a claim she has evidence for—but he argues it's "distressingly plausible" given the real pressures on the system: mounting deficits, countries moving away from US debt, gold supplanting the dollar in central-bank reserves, Japan's yen troubles, and the Treasury's genuine push (via stablecoins) to create new appetite for US debt. He uses it as a launchpad for his central thesis: that fear is how free societies get talked into handing government ever more authoritarian control, that "there's always a danger on the horizon" to justify it, and that America should deliberately run the opposite experiment—putting freedom first and being willing to pay a price for it—because politicians are downstream of what the public rewards. The conversation also covers Mayorkas's "kill-where" cyber-threat warnings and a lengthy, deliberately even-handed tangent on Elon Musk—pushing back on the "he's just a government front" framing while flagging genuine conflict-of-interest concerns and arguing the real accountability target is bias in AI systems like Grok, which can be tested and measured.

The team dives into the genuinely alarming frontier of AI security—starting with a vivid analogy: The Manchurian Candidate, the 1962 film about a sleeper agent who acts normal until a trigger phrase activates him. The host explains that this is no longer fiction for AI: a study from Anthropic's alignment team, conducted with the UK AI Security Institute and the Alan Turing Institute (described as the largest data-poisoning investigation ever), found you can plant a "backdoor" in a large language model with as few as ~250 malicious documents—and, crucially, that number appears to stay static even as models scale to billions of parameters. That means a hijack that everyone assumed would require nation-state-level control of training data can be pulled off by planting a tiny fraction (~0.00016%) of the data, simply by seeding it on the open internet where models scrape their training material. He walks through why this is so dangerous, then details a real-world attack tracked by security firm Huntress: a "malvertising" campaign ("Fake Agent") that exploited the fact that users can publish their own pages on the Claude AI platform, letting attackers serve what looked like an official Claude desktop download that actually delivered a password-stealing remote-access trojan—reportedly compromising 29 organizations in two days and pulling the fake page ~7,100 times before it was caught. From there the conversation widens into the hard strategic picture: that AI is a genuine, present-day advantage (his own team replicated 60% of a game's functionality with a non-programmer and a handful of tokens), that pulling back out of fear just lets adversaries like China race ahead, that the US risks falling behind on the energy and manufacturing needed to compete, and that the only sane path is to clearly delineate "weapons-grade" AI from civilian AI—regulating the former like nuclear material while refusing to throw the baby out with the bathwater. He closes on how to actually use AI well: not as a novice extracting answers, but as a thought partner you steer from first principles.

The team breaks down what the host calls a potentially dangerous rewriting of the entire global economic system—starting with the bloodbath in Japan, where the stock market reportedly shed a massive sum and the 10-year government bond yield hit 3% for the first time since 1996. He explains the mechanism most people miss: for decades, Japan held its interest rates artificially low with no inflation, creating the "yen carry trade"—investors borrowing cheaply in yen to buy higher-yielding assets worldwide, a hidden force that's quietly propped up global markets. Now that Japan finally has inflation, it can't defend those low rates or its weakening yen, and Treasury Secretary Scott Bessent—speaking from the G20—has been publicly pressuring the Bank of Japan to hike, telling CNBC he has information the market doesn't and that Japan will move toward a stronger yen. The host argues that unwinding the carry trade could force a global contraction, hitting anyone invested in markets or even just earning a wage, as investors sell assets to pile back into yen. He digs into why Japan is so exposed—it owns over 50% of its own debt, a reckless situation as rates climb—and explains "zombie companies" that only survive because borrowing is nearly free. Then he draws the uncomfortable parallel: the US, at roughly 130% debt-to-GDP and about to become the "buyer of last resort" for its own debt via the Treasury, is heading somewhere similar, with the central question being whether AI can generate enough real growth to outrun the debt before it implodes. The back half turns constructive: the host lays out what he sees as the only viable path—embracing AI rather than fighting it, building data centers "the right way" (crediting Ryan's approach of threading the needle so they're net-positive for the towns they're in), defining the line between "weapons-grade" and civilian AI, teaching people to use AI as a tool that empowers rather than dulls them, making real demands of an education system judged by its outputs, and letting students discharge loans in bankruptcy. A dense, sobering, but ultimately hopeful map of a genuinely precarious moment.

See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.