Unsafe AI and the Threat to America

Did Team Trump just make a Strait-of-Hormuz-sized misjudgment?

If you take the US government at its word, the latest AI models — even consumer-facing models packed with built-in safeguards — are now so dangerous we cannot risk foreigners getting anywhere near them.

(If you treat everything this particular US government says with skepticism, you may think it is simply trying to decapitate a trillion-dollar company before it goes public because it doesn't like one particular CEO.)

A necessary fight? Or a Strait of Hormuz-sized misjudgment?

The tech world, especially tech influencers on X, are debating whether the Trump administration really believes that Anthropic's Fable 5 and Mythos 5 models pose a terrifying national security threat.

As software engineer Corey Ward wrote: "Can we stop calling an LLM finding bugs in a codebase it has access to a 'jailbreak'? That's not what this is." Ward concluded his post by saying the Trump administration bringing down the hammer on Anthropic, "is entirely down to politics."

On Substack, Neal Achilson and Adam Thierer wrote: "A leading U.S. AI company was forced to take down a product that millions were using based on non-public, unexplained concerns of a few government officials. This isn’t the red-tape risk of the FDA. It’s more like the FDA demanding, out of the blue and without explanation, that everyone stop drinking milk — if milk was 50% of last year’s stock market gains."

Meanwhile, "Secretary of War" Pete Hegseth proved he sees part of his job as waging war on Americans he doesn't like by tweeting some more of the chest-thumping lies he's now famous for.

As I wrote on Saturday and Axios confirms today, it's quite likely that this gang of fools simply made a Strait-of-Hormuz-sized misjudgment when it dropped it's Friday night bomb on Anthropic, forcing it to shut down Fable and Mythos.

As Mike Zapler writes for Axios, the Trump administration's action risk sending foreign governments and companies a clear message: "Don't build your future on U.S. AI."

Already, Trump's action has created a "big moment" for China, making that country's cheaper, open-source models — which can be downloaded and run locally, free from any government threat — even more attractive to American businesses.

Whatever happens next, the situation demonstrates again that all of the Trump administration's decisions are made either by swaggering buffoons, corrupt charlatans, or weak and unprincipled bootlickers.

Who can we trust with AI?

The midterm elections are less than five months away and two things are already blindingly obvious:

1. Trump will do absolutely everything his psychopathic brain can think of to rig the elections in Republicans' favor
2. Candidates, SuperPACs, and a variety of bad actors are already using AI in new and malicious ways to poison our democratic discourse

Even before Anthropic's Mythos and Fable models, AI had already made it cheap and easy to overwhelm voters with hyper-targeted deepfake videos, fake candidate audio, synthetic “local” news, and bot-amplified rumors intended to blur reality, incite emotions and suppress turnout.

The latest models open the door to agentic election manipulation, allow users to spread lies and mistruths through relentlessly generated messaging campaigns— and to launch new kinds of cyber- and social-engineering attacks against our elections themselves.

As of now, of course, the most powerful models are available only to the government, the AI companies who developed them, and the foreign hackers who may have stolen them just as Trump "kneecapped" America's cyber defenders.

Trump warned us about "the enemy within." Maybe he just had the enemies wrong.

For years, Trump has railed against "enemies from within." He usually defines those enemies as anyone who would use the law or the ballot box to thwart his criminal schemes.

And Trump is famous for telling us that people with non-U.S. heritage — whether it's a GOP rival, a "Mexican" judge, or any non-white Democrats — cannot be trusted because of their perceived dual allegiances.

He's even told American Jews he views them all as de facto Israelis.

Now Trump tells us that Fable 5 and Mythos are so dangerous that no foreigners — even foreign nationals who helped create them — should no longer have access.

This new approach could cause massive problems throughout Silicon Valley, where 66% of tech workers are foreign born — and tens of thousands are foreign nationals working here on green cards or HI-B visas.

Meanwhile, Trump's embrace of AI has turned the South African-born election criminal Elon Musk into a trillionaire and helped German-born Peter Thiel to embed his Orwellian technologies throughout the U.S. government, especially in areas such as defense, immigration enforcement and healthcare.

Many other tech companies whose revenues depend heavily on global markets are headed by foreign-born CEOs, including tech giants such as Alphabet and Microsoft, where both CEOs were born in India, chip giants Nvidia and AMD, whose CEOs were born in Taiwan, and Intel, whose CEO was born in Malaysia. In 2025, those five companies generated more than $400 billion in revenue from non-U.S. markets.

I'm not saying that foreign-born tech CEOs are more beholden to their countries of birth than they are beholden to America.

But I'm pretty sure they're more beholden to their shareholders than they are to their employees or to the citizens of any country in which they do business.

And I know for sure Musk and Thiel are in a separate category.

They're not just using their money and influence to corrupt and manipulate Trump, JD Vance, and others.

They're also actively campaigning for the end of electoral democracy.

Thanks for reading and sharing! And an extra special THANK YOU to all my paid subscribers, past and present, and contributors to my "coffee fund." While paid subscriptions are always welcome, I’m continuing to offer all content free, so choosing a “free” subscription means you'll never miss an issue of UNPRECEDENTED.

To support this newsletter without subscribing, consider a one-time tip at Buy Me a Coffee.