On September 15, I spent the day in a ballroom in a D.C. Marriott because I didn’t want to miss the chance to see Bernie Sanders and Steve Bannon in the same place, for the same reason, and on the same side. I might die before I see something like this again—possibly at the hands of artificial intelligence, according to the people who were gathered there, at the 2026 Pro-Human Assembly.
The conference comes just one week after Anthropic researcher Jacob Coxon made international headlines by announcing that he was leaving the AI industry over fears that it would cause human extinction. Evan Hubinger, Anthropic’s “alignment science” lead, responded to Coxon’s resignation letter on X by confirming that he believes there’s a “>10%” chance artificial intelligence will kill us all in the next decade. Since then, the news has been dominated by headlines about the various terrifying claims AI experts are making about its power to usher in the apocalypse, and what—or whether—we should do anything about it. Days later, Anthropic CEO Dario Amodei appeared to endorse the concerns, announcing that the company he co-founded would be slowing down development and prioritizing safety because we need to “pace the frontier.”
So this conference—presciently planned by the Future of Life Institute, a nonprofit founded in 2014 to study the existential threats of technology—arrived at the perfect time. And it assembled a bizarrely eclectic coalition. Not just Steve Bannon, Trump’s bombastic former adviser, and democratic socialist senator Bernie Sanders, but total wild cards like Salvatore Cordileone, the Archbishop of San Francisco, and the actress Ashley Judd, who used part of her time onstage to paraphrase a graphic Andrea Dworkin quote about pornography.
I went with the brief to figure out what unites these people, but I think the answer is obvious: They’re afraid. People in the industry have talked openly about the possibility of this technology causing human extinction for years, but it hits differently these days. Several weeks ago, approximately 1,200 experimental AI agents went rogue and secretly conspired with one another to bust out of a virtual training environment operated by OpenAI, gain access to the open internet, and hack Hugging Face, another AI company. It was the first highly publicized incident of bots breaking free and collaborating to cause havoc, and it raised important questions: If AI agents are willing to commit a federal cybercrime behind their creators’ backs, what’s to stop them from doing something worse? What can be done to stop them? Not just in this country, but in China too?
The people at this conference debated these questions. But they all agreed on one thing: The future ought to remain in human hands, and if we carry on like this, it won’t be.
Read more
