AI Doom Is Misguided, Not a Psyop

Did you see the news? Everybody is freaking out about artificial intelligence, which poses a significant likelihood of killing everyone on the planet once it achieves superintelligence…at least according to a bunch of tech folks who might have their own reasons for stoking mass panic.

This media frenzy has led to blowback from a bunch of people, including Elon Musk, who say AI doomers are engaged in a “psyop”—in other words, they are using inauthentic means to manufacture a nonexistent controversy over AI. Proponents of the psyop theory point to a high degree of coordination among the doomers, though they are glossing over some less conspiratorial explanations for why politicians and reporters are boosting their claims.

Ex-Anthropic employee Jacob Coxon triggered this meltdown earlier this week when he announced he was leaving the company due to his own concerns that Anthropic and OpenAI—where he previously worked—were racing toward superintelligence without sufficient concern for how an advanced AI could threaten humanity. His thoughts on X went super viral and were widely disseminated by other tech people, politicians, and commentators.

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

— Jacob Coxon (@hilbertspaess) September 9, 2026

You might have expected Anthropic to push back on his claim, but one employee, Evan Hubinger, actually chimed in to echo Coxon’s concerns.

“We really do earnestly believe AI could kill all humans!” he wrote. “I personally think it is >10% within the next decade.”

Jacob is correct here—we really do earnestly believe AI could kill all humans! 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. https://t.co/QAIHiFP3QZ

— Evan Hubinger (@EvanHub) September 9, 2026

A greater than 10 percent chance of something happening is significant. I think most people would not want the AI companies to proceed with their plans if the risk of global extinction was really that high. Note that Hubinger also explicitly rejects the notion that this is a “marketing stunt,” which is an accusation frequently leveled at AI doomers: namely, that they are trying to drum up interest in the product by making it sound dangerously cool, which will help to increase its appeal to and ultimately its value.

That said, it still could be a marketing stunt! There’s like a 10 percent chance our totally awesome product goes nuts—isn’t that amazing and scary? does in fact sound like a marketing gimmick rather than a sober and scientifically reasoned analysis of the likelihood that superintelligence will produce catastrophe. How was the 10 percent figure arrived at? Did they weigh the odds of the superintelligence saving millions of lives? Or counteracting a foreign adversary’s malicious AI? If the people working on these systems really think they’re dangerous, they have a responsibility to be more specific about the kinds of threats we may face.

Vague predictions of tech-related doom are more than sufficient to make the public extremely afraid, however. Coxon’s post was shared everywhere, and he was quickly summoned onto cable news shows. Op-eds appeared in newspapers and magazines. Regulators signaled their readiness to spring into action. Anthropic itself seemed ready for the news blitz, and issued statements begging the federal government to do something to constrain not just its AI—but also its competitors’.

This sort of behavior is nothing new in the tech world. Meta, for instance, has routinely issued public pronouncements in support of greater regulation of social media companies; in years past, critics have argued that the legislation they support could wind up helping Meta’s products by solidifying Facebook and Instagram as market leaders. More recently, following Meta’s decision to settle various lawsuits for billions of dollars, the company said it was in full of support of government-mandated childproofing—so long as the government reined in YouTube and TikTok, too.

But if Anthropic’s decision to align with AI doomers is cynical and self-serving, it’s still not a “psyop,” or an elaborate conspiracy to trick the American people into supporting more regulation—something critics of the doomer position are also alleging. The truth is that neither politicians nor the mainstream media need much of an excuse to succumb to knee-jerk anti-tech thinking: They default toward wanting to prevent alleged harms.

Who is held accountable when AI goes rogue?

I’m launching an investigation into OpenAI – Americans deserve answers pic.twitter.com/bL9zHuHhH6

— Josh Hawley (@HawleyMO) September 10, 2026

Both the media and the government are biased in favor of action: Someone should do something. And AI is a big something. We’ve all been raised on science fiction stories that predicted killer robots would try to take over the world: Terminator, The Matrix, Dune, Battlestar Galactica, and so on. It’s easy to see why flashy headlines warning that tech insiders not-so-secretly believe a rogue AI is only a few years away from destroying humanity is going to sell a lot of newspapers. Similarly, no politician wants to be seen as being insufficiently attentive to this problem. Even Sen. Ted Cruz (R–Texas), a comparatively pro-tech political figure, prefaced his recent remarks on AI by first stressing his bipartisan do-something approach to tech legislation.

So it’s not really a surprise that the world of political commentary is buying what that ex-Anthropic whistleblower is selling. The last thing this debate needs is for the pro-AI side to succumb to the paranoia that broadly characterizes the doomer side. It’s not a psyop; it’s a mistake. (And no, that sentence was not written by AI. I can use this construction too, ChatGPT!)

We discuss President Donald Trump’s proposal to give every single American a $5,000 “dividend” check if the Republicans somehow win the midterms.

So far, I am extremely happy with the footage I’m seeing from the Legend of Zelda: Ocarina of Time remake. How about you?

The post AI Doom Is Misguided, Not a Psyop appeared first on Reason Magazine.