Hoax

Amodei skepticism: “The calls by four American tech CEOs for a slowdown in AI development is a big economic and political moment, though to what end isn’t clear,” writes The Wall Street Journal‘s editorial board, responding to the Dario Amodei letter published over the weekend that called for an AI development slowdown, which was quickly agreed to by the heads of other major AI companies.

“The CEOs can also see how the politics of AI are moving toward panic. The opposition to data centers based on misinformation is one signal. They can also see the plaintiff bar circling in wait for an incident when rogue agents do more harm than the Hugging Face event,” continues the Journal, referring to incidents in which swarms of rogue agents have attacked other companies and stolen their data. “The warnings of Apocalypse are also growing, often uninformed and amplified by those like Bernie Sanders who want government to stop the industry until the politicians can run it. By offering to slow down on their own, and invite outside monitors to inspect their models, the CEOs are hoping to head off such an outcome.”

Crucially, writes the editorial board, “nothing now is stopping these firms from ‘pacing the frontier’ on their own, to use Mr. Amodei’s phrase. OpenAI and Anthropic are leading the frontier. If what they see in their shops is truly dangerous, by all means be responsible and ‘align’ development. Slow down on your own,” they add. “The complication comes when these firms seek outside help to aid their cause.”

What might this look like? Regulatory capture is one risk. Figuring out how to pace the American frontier, given the growing threat posed by Chinese AI, is another. (Though it’s worth noting that the Chinese Communist Party is even more sensitive to threats that could unseat its own power, and to the free flow of information, so it’s not clear to me what the future of Chinese AI looks like, specifically. Probably government control of AI for espionage purposes.)

Some have theorized that part of the reason why this call is coming now is that Anthropic (and perhaps OpenAI as well) has no buzzy new model in the works that will be released soon; if you brand a slowdown as one for safety and prudence, that’s more appealing than if the pace of innovation slows. Note that everyone’s trying to go public soon, so it’s very hard to tell what’s real and what’s not.

That said, our own president seems to believe this is all a “hoax.” In an onstage call yesterday with chipmaker Nvidia’s CEO, hosted by the All In podcast, President Donald Trump told Jensen Huang very decisively that “the robots will not be taking over.”

“America is able to lead and to do it safely,” replied Huang.

Trump then clarified his thoughts via Truth Social: “AI taking over the World, destroying Humanity, and all other things bad, is a HOAX.”

And: “The AI Hoax being perpetrated by the Radical Left Dumocrats is reminiscent of their Global Warming Scam of not so long ago, where everyone was going to die from extreme heat. What happened?”

“I am the Hoax Buster,” he continued, “and I’m right now breaking another Hoax — That AI is going to take over, consume, and destroy the World, and that Robots will be marching into our Cities, and getting rid of us all! This is even wilder than the RUSSIA, RUSSIA, RUSSIA HOAX, or the Global Warming Scam. Thank you for your attention to this matter!”

I don’t think that Trump is paying very much attention, though I suppose it’s good that he’s broadly pro–data center. The concern has not been that robots will march into our cities; sneakier cyberattacks and ramped-up espionage are my concerns, as well as attacks on critical infrastructure. Recursive self-improvement—AI agents designing and coding the next generation of successor agents—creating problems (i.e., “misalignment” from creator objectives) is one major worry. It’s also possible that things like financial scams become much more sophisticated and successful, or that we have massive leaks of private documents and information. What happens in the future if information is no longer secure, and sensitive documents can be easily and convincingly duped?

It’s not clear that regulation is the answer, per se, or that an industrywide pause prevents these problems from ultimately emerging. It’s also not clear Trump understands even a fraction of the problem.

Scenes from New York: I’m intrigued by Zoomer faux-nostalgia, yearning for a time of which they were never a part.

America’s golden era was NYC pre 9/11.

Pictured is the “Windows on the World” restaurant in the North Twin tower

At first I thought these pictures were from one one of those 90s high society vibe ai generated picture accounts but no pic.twitter.com/M5ZtUmxfvm

— matador (@matadormurderer) September 11, 2026

QUICK HITS

• Researchers Nicholas Carnes and Noam Lupu just published a political science paper on politicians’ oft-exaggerated working-class bona fides and how actual “people from working-class jobs almost never go on to hold elected political office.” Though “52 percent of United States citizens hold working-class jobs, only 2 percent of state legislators in 2023 and 2 percent of members of Congress in 2019 ‘held working-class jobs prior to holding public office,'” notes The Washington Post in a write-up of the paper.
• “About 15 years ago, executives from U.S. fruit grower Driscoll’s went to China with a goal of turning blueberries into that country’s next big food trend. The California company dispatched one of its experts to scout farmland in southwestern China’s Yunnan province, contracted with global blueberry-genetics firms to lock up the best varieties and brought in high-tech growing technologies to mass produce a fruit few Chinese people then ate. By 2020, it was cranking out thousands of tons of blueberries for Chinese supermarkets,” reports The Wall Street Journal. “As tales of blueberry fortunes spread, Chinese entrepreneurs rapidly built farms of their own. They adopted high-tech growing techniques similar to Driscoll’s, even allegedly copying greenhouse designs used by farmers growing for Driscoll’s. They also swiped the company’s patent-protected varieties, a Chinese court has ruled.”
• Leveling it all:

Remember!

If the metrics and tests show that your preferred education policies aren’t working, you can always just abolish the metrics and tests. https://t.co/vwnu2d9fGB

— Jeremiah Johnson 🌐 (@JeremiahDJohns) September 14, 2026

Q: Do you expect Congress would need to approve the $5,000–

TRUMP: I don’t know, but it’s easy enough. It’s $5,000 to all adults in the country, and we can easily handle that because we’re taking in so much money pic.twitter.com/PmSgSgibAq

— Aaron Rupar (@atrupar) September 13, 2026