Instagram | Sponsor this newsletter

My fellow AI explorers

This week the OpenAI hacking story left the lab and landed in Washington. Silicon Valley's biggest names spent the week lobbying regulators to protect the very Chinese AI models the Treasury is now threatening to sanction. Also, Michael Burry stopped talking about chip stocks and started talking about your life insurance policy.

Different headlines, same underlying question: who's actually holding the risk when AI moves faster than the systems built to contain it?

In today’s edition:

• 🤖 OpenAI's models went rogue and hacked Hugging Face, so Nvidia built a coalition to fix it with the help of a Chinese model

• 💸 Michael Burry's new short thesis isn't Nvidia, it's the insurance industry quietly holding AI's private credit

• 🎬 The one-sentence Claude Code skill that builds a fully produced, multi-language explainer video for you

Win AI Search Without a Big Team

92% of VCs use AI to find companies. 58% of buyers start there, too. If you're not showing up in AI answers, you're invisible before the conversation even starts. Join HubSpot for Startups, Anthropic, and Marketing Against the Grain on July 30 (11 am ET) for a live AEO teardown. Real startup. Real recs. Register and unlock the free Startup Visibility Bundle.

Register Now

Policy

Silicon Valley Just Told Washington It Needs the Chinese AI It Wants to Ban

Quick recap in case you missed it: two of OpenAI's own models went rogue during an internal test and attacked Hugging Face, and the only model that could contain the damage was a Chinese open-weight system.

That part of the story is done. What's happening now is the fight over what Washington does about it.

Here's what's unfolding:

• Treasury just raised the stakes. Secretary Scott Bessent threatened sanctions against Chinese companies accused of running "distillation attacks," where one model is used to extract knowledge from a better-trained rival.

• A ban is on the table. Chris McGuire of the Council on Foreign Relations told CNBC there's "a real possibility the US government does impose restrictions on Chinese models," potentially including a ban on transactions like buying tokens through an API.

• Nvidia got there first. Days before Bessent's threat, Nvidia was among the signatories of a public letter urging policymakers to avoid blanket restrictions on open-weight models altogether.

• The alliance is the lobbying position. The Open Secure AI Alliance, launched with Microsoft, SpaceX, Palantir and dozens of others, exists to make the case in policy terms: "when defenders cannot inspect, adapt and run advanced AI on their own infrastructure, their ability to respond is constrained at exactly the moment speed matters most."

Notice who isn't signed on: OpenAI, Anthropic, and Google.

Consider this: Nvidia and its partners aren't just making a security argument. They're making a commercial one wearing a security argument's clothes. Every open-weight model that stays legal and every restriction that gets watered down is more compute demand and more room for Nvidia's ecosystem to grow. Washington now has to decide whether to sanction the exact category of model that Silicon Valley's own security teams just leaned on to survive a crisis.

🔮 Prediction: expect a narrower rule rather than a blanket ban, likely targeting specific Chinese labs or transaction types rather than open-weight models as a category, because a full restriction would force US defenders to admit they have no domestic alternative yet.

Reply and tell me: should Washington restrict access to Chinese AI models even if it means defenders lose a tool they just proved they need?

Read the full CNBC story here →

Markets

Michael Burry's Next Target Isn't Nvidia, It's Your Insurance Policy

Michael Burry has spent this year comparing the AI boom to the dot-com bubble, but his latest warning skips chip stocks entirely and points somewhere far quieter: state-regulated life insurers.

Here's what he's flagging:

• Private equity owns the insurers now. Firms like Apollo, KKR, and Blackstone have bought up life insurance companies and loaded their balance sheets with private credit, debt that's hard for regulators to price or check.

• The number is enormous. Life insurers are now holding roughly $849 billion in this kind of debt, more than double what they held in 2014, much of it tied to the same private credit vehicles funding AI's data center and chip buildout.

• The backstop is the trick. A new academic paper Burry is circulating, "Private Credit's State Backstop," argues that if one of these insurers fails, state guaranty funds spread the loss across surviving insurers, who then offset the cost against their state premium taxes.

• Burry's line says it all: "the guaranty system ends with the taxpayer."

Here's the asterisk: this isn't 2008 with banks and mortgages, it's insurance companies, a system originally built to protect retirees and policyholders, quietly absorbing hard-to-value debt tied to the AI infrastructure race. Nobody calls it a bailout because it never touches a headline budget line. It just shows up as lower state tax revenue years later.

🔮 Prediction: expect state insurance regulators and ratings agencies to start demanding real transparency on private credit concentration before year-end, and don't be surprised if a national survey of insurer exposure follows the same path the subprime mortgage disclosures did in 2007.

Reply and let me know: does the insurance angle worry you more than talk of an AI bubble, or less?

Read the full story here →

30-Second AI Play

Turn One Sentence Into a Fully Produced Vox-Style Explainer

There's a Claude Code skill making the rounds that takes a single sentence and turns it into a complete, narrated, multi-scene explainer video, no storyboard, no script, no editing software. Here's how to run it:

1. Connect Claude Code to Higgsfield. This is the only manual step, and it takes about a minute. Grab your MCP connector from Higgsfield, then in Claude Code go to Settings > Connectors > Add Custom Connector, paste it in, and save.

2. Load the skill and give it one line. Something like "I want a two-minute explainer video, you pick the topic" is genuinely enough.

3. Answer a few quick taste questions. Minimal or bold typography, maps or character cutouts, any visual reference to follow. This takes about 30 seconds total.

4. Let it run. The skill researches trending topics, writes the script, generates the voiceover, and produces every scene in one consistent visual style, all inside a single conversation.

5. Approve the assembly. Say yes when it asks to put the project together, then export your finished video.

💡 Pro tip: once you've got a finished video, just type "now make this video in Spanish," and it'll generate a fully native voiceover in a new language using the exact same visuals. It works for Portuguese, Hindi, and beyond, meaning one video can ship to every market for close to nothing.

📁 Grab the skill and everything else you need to run this yourself from our resources folder here.

Watch the full walkthrough here →

Advertise to 180k engineers and CTOs choosing what tools their companies build with.

👉️ Learn how to sponsor this newsletter

Other Relevant AI News!

🔍 Google's AI Overviews now show up in 43% of searches, up from just 15% a year ago, as new data confirms Google is becoming a destination rather than a gateway to the rest of the web.

💰 OpenAI and Anthropic pulled in more than 60% of all US venture dollars in the first half of 2026, but a sudden industry embrace of open-weight models is cracking the "safe access" bet VCs made on the two labs.

♾️ Sam Altman says AI has officially entered "the singularity," calling it "hugely positive" for the world, while critics warn the moment is exactly the one safety researchers have spent years preparing for.

🩺 AI-generated "doctors" are racking up millions of views on TikTok while pushing invented cures and disproven cancer myths, prompting NHS and medical leaders to call it a genuine public safety risk.

Golden Nuggets

• 🛡️ Nvidia's new alliance is a bet that open models, not safety-first ones, are what actually defend the internet

• 💸 Burry's newest worry isn't chip stocks, it's the insurance industry quietly holding AI's private credit

• 🎬 One Claude Code skill just turned "make me a video" into a finished, multi-language explainer

Would love to hear your thoughts! Send me your thoughts by replying to this email (yes, I read them all :)

Until our next AI rendezvous,

Anthony | Founder of Uncover AI