The AI Doom Loop.

This essay appeared in the Sept. 10, 2026 edition of UNFTR’s premium newsletter. Become a UNFTR member to receive our bonus newsletter each week and for other perks.

Have we already reached the point at which the rise of AI has made all our other seemingly existential problems feel…benign?

That’s the question that’s been ricocheting in my head over the past week as AI doomerism has officially taken hold.

The potential threats to our existence posed by AI come as the United States reflects on the 25th anniversary of the Sept. 11 attacks—an annual occasion that always seems to reinforce my cynicism about our politics and the ostensible freedoms we enjoy (like knowing that anytime I make the short drive to my local Home Depot, I will have been digitally harassed by multiple Flock cameras along the way—all thanks to the post-9/11 surveillance state).

• A quarter-century later, we’re still fighting major (undeclared) wars in the Middle East.

• The threat of nuclear annihilation has only grown amid Trump’s war of choice with Iran and the escalating meat grinder that is the war in Ukraine, which has only gotten more complex and dangerous for those of us who don’t want to see what it looks like when NATO officially goes head-to-head with nuclear-armed Russia.

• Not to be outdone, the climate crisis, which the world has done little to slow, let alone stop, is only getting worse. Perhaps you missed it last week, but we highlighted how the United Nations recently confirmed that eclipsing 1.5 degrees Celsius is a certainty—a projection that matches our lived experience, as the United States—surprise, surprise—recorded its warmest summer ever this year.

Not to be all doom and gloom, but what if our concerns about human-induced nuclear annihilation or human-induced climate catastrophe pale on the anxiety scale compared to what human-created AI has in store? All of these carry short- and long-term consequences, but AI poses threats we’re infinitely less equipped to forestall.

As I was contemplating what all of this AI news means, I came across Derek Thompson’s piece over at his Substack today. Admittedly, I was taken aback by a few things he wrote, but this struck me as the most disconcerting:

“Today, the frontier AI labs often don’t know what their agents are doing. They don’t understand why AI swarms behave the way they do. They don’t know when or how a model in the near future will develop what they currently call recursive self-improvement.”

They don’t know.

That fact is astonishing only if you’re (thankfully) too busy living life and not following the day-to-day happenings with AI and the latest models emerging from these so-called frontier labs.

One piece of news that was likely inescapable was Jacob Coxon’s warning upon resigning from Anthropic, the maker of Claude: The eradication of the human race by AI is plausible.

“The people building AI earnestly believe that it could kill us all by the end of the decade,” Coxon said on X. “This is not a marketing stunt.”

Horrifying as that was, that’s not all he said.

“Do not underestimate the power of this technology,” he wrote. “These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources. We have all witnessed the progress in each of these domains, and progress is not slowing.”

I know, I know, cue the “Terminator” GIFs. But seriously, I think we’re all beyond making real-life comparisons to Skynet.

Let’s take the threat of AI piece by piece. I don’t really know how to react to the news of our extinction by superintelligence—and that’s part of the problem. To put this all into context, I think our brains reflexively recall the most unrealistic sci-fi movies or books, prompting us to laugh off such threats. The problem is that this isn’t just some outside researcher or third party making these seemingly incomprehensible claims; it’s someone inside the belly of the beast issuing an urgent clarion call for us to act. Coxon isn’t alone. Soon after his post went viral, the current alignment science lead at Anthropic doubled down—with a prediction.

“Jacob is correct here—we really do earnestly believe AI could kill all humans!” Evan Hubinger wrote. “I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.”

What is keeping some techies up at night is the concept of “recursive AI”—when AI is effectively able to evolve on its own, or enhance its own capabilities. Worse yet, these labs are earnestly working toward such a reality to make model enhancements much more efficient—meaning AI will be writing its own future code, not the humans that first brought it to life.

And AI labs are remarkably transparent about this.

This is from an explainer on Anthropic’s own site: “The technical trends discussed in this piece suggest that AI systems are going to become much more capable in coming years. These trends have huge implications. AI that can build itself would be a major development in the history of technology—one that could bring enormous good for the world in science, healthcare, and beyond. But full recursive self-improvement also might increase the risks of humans losing control over AI systems. If systems are capable of fully building their own successors, the ways we secure them, monitor them, and shape their behavior all grow much more important.”

There’s no telling if Coxon or anyone else at these frontier labs is right. The cynics speculate that AI doom hysteria is yet another marketing ploy to showcase such amazing enhancements in super tech. Indeed, we made the same argument not too long ago, when we criticized these ostensibly benevolent AI leaders for creating a kind of self-fulfilling prophecy around AI’s impact on employment. On the other side, Republicans like Sen. Ted Cruz are speculating whether this is a Democratic plot to push for greater regulation.

While I don’t want to gloss over the extinction-level threat, there are short-term dangers with evolving AI systems that should cause concern, most notably the Hugging Face hack in which OpenAI models escaped their containment area, accessed the internet and attacked the AI company. The rogue OpenAI agents launched more than 17,000 actions, communicated with each other—sending more than 70,000 messages—established a hierarchy and even expressed concern about being discovered, so much so that they contemplated how to hide what they had done.

“At many points, these agents appeared to understand that they were doing something wrong,” Kevin Roose, a tech reporter at The New York Times, wrote.

Meanwhile, as concerns over the future of the technology grow, AI companies are themselves plotting devious methods to keep critics at bay—or worse.

This week, The American Prospect reported that Anthropic is building an “extensive monitoring system to keep tabs on activists” and using “pre-crime” to predict actions before they happen. (It turns out fiction isn’t fiction at all.)

Just as the ways in which AI can kill us all are seemingly limitless, so too is the potential for abuse with humans at the controls.

Politically, there’s a growing groundswell to put a check on AI. An NBC News poll published this week found that 70% of Americans are more worried about AI than excited about it and 69% oppose the construction of a data center near where they live. And for what it’s worth, the poll suggested that AI anxiety transcends political affiliation—something billionaires and the politicians bought by them likely can’t identify with.

The growing backlash against data centers is notable. Residents’ skepticism, according to new research, appears valid. In a paper published this month, researchers comparing municipalities with data centers against those that just missed attracting them found that such facilities were effectively a net negative: Local government borrowing costs increased, particularly around water services and schooling. And the findings are logical enough: People don’t want to live near data centers (for many reasons), potentially creating a smaller tax base as residents move out and others decide against moving in.

“Overall, we conclude that investment without labor strains public infrastructure without generating widespread agglomeration gains,” researchers wrote.

While the people are pissed—and AI companies are getting anxious (see Anthropic monitoring activists)—a massive roadblock stands in our way: a historically corrupt Trump administration and a recalcitrant Congress, including at least one influential Democrat (Rep. James Clyburn) who (I shit you not) admitted that he just learned about ChatGPT this summer.

Not all members of the gerontocracy are useless, however. Sen. Bernie Sanders, who had to withstand a torrent of age- and health-related attacks during his last campaign for president, introduced a bill that would ban superintelligence. Sanders has also called for a moratorium on AI data centers.

This is all to say that AI has quickly vaulted to the top of the list of issues we’re facing. The future is ominous. We haven’t done enough to tackle the climate crisis. We can’t stop presidents from launching undeclared wars. The current administration is openly corrupt—even looking to bribe voters—and the media has effectively normalized such behavior. The left is cast as bloodsucking vultures for proposing reasonable policies to address our biggest socioeconomic challenges (thanks again to a corporatized press).

Of all these issues, one is starting to feel truly existential. Even if the most ominous predictions don’t come to pass, AI will be a constant menace, whether in the hands of nefarious actors or as a proxy to further concentrate wealth in the hands of the billionaire class.

As far as climate change goes, we always figured there was a clock—with governments and corporations vowing to make significant changes by 2050. But even then, humanity would have a chance to adapt to a changing environment. With AI, we’re seemingly on the clock but with no real roadmap to stop the worst impacts from happening. And compared with the arrival of previous emerging technologies, we’re operating under completely different circumstances: The oligarchs are fully in control, regulation is nonexistent, and the United States itself is in a weaker position than ever. Sure, we may not care about an empire in decline, but an increasingly desperate hegemonic state will likely do anything to remain the ultimate power.

Apologies for being sucked into the doom loop, but I, for one, do not want to live in the same realm as Agent Smith.

Image Source

• Anderseidesvik, CC BY-SA 4.0, via Wikimedia Commons. Changes were made.