Trump’s Dangerous AI Nationalism
Plus: TikTok’s pro-Israel future; Trump’s neocon conversion; Eurovision impairment; and more!
This week, as the United Nations General Assembly convened in New York, the UN christened two new vehicles for advancing international discourse on artificial intelligence: the Independent International Scientific Panel on AI and the Global Dialogue on AI Governance. A UN statement said they were “designed to bring effective global AI governance a step closer to reality.”
Well, that’s something. On the other hand, it would have been hard to move effective global AI governance any further away from reality than it is. The United States, one of the world’s two AI superpowers, is led by a president who rejects the whole idea of global governance (which he calls a form of “coercion and dominance”).
And this week, for good measure, his top science adviser issued a statement trashing this particular application of the idea. “We totally reject all efforts by international bodies to assert centralized control and global governance of AI,” wrote Michael Kratsios, head of the Office of Science and Technology Policy. The US, he said, would pursue things like “accelerating AI innovation, building AI infrastructure” and “establishing American AI as the global gold standard” and safeguarding “the independence and sovereignty of nations.”
Kratsios didn’t explain how fiercely protected sovereignty would help America escape the various kinds of AI-abetted disasters that are international in nature. Like, for example: a massively lethal global pandemic caused by a novel bioweapon made abroad with AI’s help; or a whole new kind of autonomous, self-replicating, and infinitely adaptive cyberthreat; or the US-China war that could result from a headlong race to “superintelligence”—a race that, as the finish line approaches, strongly incentivizes whichever nation is in second place to cripple the other nation’s AI development by any means necessary, including “kinetic” means; or various other bad things, like the catastrophic social destabilization that could result from accelerating AI advance—acceleration that is effectively unstoppable in the absence of an agreement among the US, China, and other nations.
You can’t really call Trump’s rejection of international AI governance a disappointment. There was never much hope that win-win responses to non-zero-sum dynamics would be embraced by a president who once wrote, in a book titled Think Big and Kick Ass, “You hear lots of people say that a great deal is when both sides win. That is a bunch of crap. In a great deal you win—not the other side. You crush the opponent and come away with something better for yourself.”
What is more disappointing than Trump’s attitude is the attitude of people who should know better. With the executive branch constitutionally numb to the nature of the AI policy challenge, hopes of rousing the legislative branch are dimmed by the silence of smart, knowledgeable potentially influential people in the AI community.
I’m not just talking about the obvious examples—corporate AI players, like Sam Altman, who used to advocate international regulation back before they realized that fostering an arms race mentality by fanning Chinaphobia was a good way to keep regulators at bay and maybe land a few government subsidies. Even many less obviously compromised players—even people who declare themselves members of the “AI Safety” community—are saying little if anything about international governance and are singing along with Sam about the need to “beat China.”
There are a few exceptions—including, significantly, Geoffrey Hinton and Yoshua Bengio, who are more responsible for the “deep learning” revolution that birthed every kind of AI currently in the headlines than any two other researchers in the world (and who shared the 2018 Turing Award, computer science’s Nobel, along with Yann LeCun of Meta in recognition of this role). Their names, along with those of 10 Nobel laureates and more than 200 other notables, were affixed to a letter released at the UN this week calling for “an international agreement on clear and verifiable red lines” around AI.
Another exception is David Krueger, an AI researcher and safety advocate who signed that letter and appeared on the episode of the NonZero podcast that we posted today. When I asked him about this— why so few AI safety researchers are talking about international regulation and why so many sound like China hawks—he said, “Five or ten years ago, everyone agreed that we needed international cooperation—everyone I talked to in the safety community. And then I think people went to Washington and started playing insider politics.”
Krueger added that “some people genuinely think that it’s really important the US beat China,” but others are acting out of “political opportunism” and some are just “scared to disagree with other people in the community and they’re scared to disagree with the powers that be.” Krueger believes an AI race with China is “completely crazy”—but saying that is “just not really in the Overton window of mainstream politics in the United States at the moment.”
Krueger and I also discussed the new book co-authored by AI doomer-in-chief Eliezer Yudkowsky, If Anyone Builds It, Everyone Dies, which hit the New York Times bestseller list this week. And we compared Yudkowsky’s dramatic “AI takeover” scenario to a scenario explored in a paper called “Gradual Disempowerment” that Krueger recently co-authored: an incremental but persistent and ultimately dismal surrender of agency to AI.
The Yudkowsy scenario is sometimes called the “loss of control” scenario, but the truth is that every grim scenario, including AI-abetted pandemics and a socially destabilizing pace of job displacement, is a loss of control scenario: If these things happen, that means we’ve failed to control technology adequately. As UN Secretary General Antonio Guterres put it this week, “the question is whether we will govern this transformation together or let it govern us.”
—RW
This week, as President Trump signed an executive order that will let a group of American investors take over TikTok’s US operation, The American Conservative ran a piece saying that the real winner will be “Israel and its lobby.” TAC notes the deal will put TikTok America’s future in the hands of players that range from reliably pro-Israel to ardently pro-Israel, such as the investment firm Andreessen-Horowitz and Oracle’s Larry Ellison. Two possible investors who have been mentioned since the TAC piece came out—Rupert Murdoch and Michael Dell—also fit that description.
In response to this analysis, NonZero’s TikTok special ops unit (“Reel Team 6”) swung into action. We figured we’d better get any content that might run afoul of TikTok’s future speech police out there now. So, though the episode of the NonZero podcast featuring Middle East expert Rob Malley won’t be posted until next week, you’ll find a TikTok extract from it below.
Malley, who was on the Clinton administration’s team at the Camp David talks of 2000 and also served in the Obama and Biden administrations, has co-authored a new book with Hussein Agha, who was on the Palestinian team at Camp David, called Tomorrow Is Yesterday: Life, Death, and the Pursuit of Peace in Israel/Palestine. In the clip below, Malley addresses the question of whether this week’s recognition of a Palestinian state by France, Britain, Canada, and Australia will make a difference.
After creating the clip below, we took a closer look at the TAC article and realized that we may have acted too late—by, like, months. The article says that, since last year, when US politicians started complaining about a “pro-Hamas” bias on TikTok, the company has hired as its hate speech manager an “IDF instructor and self-declared ‘passionate Zionist’”—and has forged a content moderation partnership with an NGO that is “closely linked to the Israeli government and staffed by Israeli spies.” Which makes it only more important that you click on this short video, thus helping us counter any algorithmic headwinds it may face:
Tiktok failed to load.Enable 3rd party cookies or use another browser
Seems like only yesterday President Trump was championing a counter-establishment foreign policy that would eschew regime change operations and forever wars in faraway lands and would abandon the vestigial Cold War reflexes that led to NATO expansion and got President Biden invested in fighting Russia to the last Ukrainian. Or was it last week?
Anyway, this week is a different story. Over the past seven days: (1) Sources told the New York Times that (as NZN suggested the day before the NYT piece came out) the bombing of alleged drug-running boats from Venezuela is actually about engineering regime change in Venezuela; (2) Trump said he wants to send a contingent of troops back to Afghanistan, notwithstanding the Afghan government’s rejection of that idea; (3) Trump said that Ukraine, which he’d previously said must accept a loss of territory for the sake of peace because it’s playing such a weak hand, is actually playing a strong hand and can regain all its territory and maybe even seize some Russian territory—and the US will be happy to keep supplying the weapons so long as other NATO members pay for them.
Did Trump have a neocon conversion experience? It’s not impossible. Former neocon (and, perhaps, current crypto-neocon) Marco Rubio, Trump’s secretary of state and a long-time fan of Latin American regime change, presumably helped steer Trump toward the Venezuela policy and maybe toward the other two shifts. Of course, Trump’s conversions aren’t known for their staying power. So tune in next week. Meanwhile, some background on the Venezuela and Afghanistan fronts:
—A bill drafted by Republicans would give President Trump sweeping authority to take military action against drug cartels he deems “narco-terrorists” and any nations harboring them, the New York Times reports. The bill could offer legal cover for future instances of the recent missile strikes on Venezuelan boats that Trump claimed were carrying drugs—and for any number of other adventures. “It’s insanely broad,” Harvard law professor Jack Goldsmith, a senior Justice Department official under George W. Bush, told the Times. “This is an open-ended war authorization against an untold number of countries, organizations and persons.” (To hear Goldsmith discuss Trump’s broader aspiration to expand executive power, check out this episode of the NonZero podcast from February.)
—The Trump administration has been talking to the Taliban about stationing US troops at Bagram Air Base for potential counterterrorism operations, the Wall Street Journal reports. A senior official in Afghanistan’s foreign ministry posted on X that a renewed US troop presence isn’t acceptable but called for cordial political and economic relations with the US. Soon after, Trump warned on Truth Social that “BAD THINGS ARE GOING TO HAPPEN!!!” if Afghanistan doesn’t relent on Bagram and told journalists that one virtue of the base is that “it’s an hour away from where China makes its nuclear weapons” (not by car: the distance is 1,200 miles). The Journal says this initiative reflects the mindset of such Trump administration officials as UN ambassador and erstwhile national security adviser Mike Waltz, who in a 2021 column for Military Times described the US presence in Afghanistan as a “strategic foothold” in the “back yard of America’s greatest rivals.”
Last week we had a Zoom call with NZN members, and dozens of them set a new endurance record—they stayed on the call for more than two hours, enduring my answers to their questions. If you’re a member and couldn’t make it, you can listen to a recording of the conversation on the paid-subscriber podcast feed. We talked about a lot of stuff, including movement building and community building.
On that second front, NZN team members Nikita Petrov and Danny Fenster, who were in attendance, would like some input: (1) Nikita invites any NZN members who want to have one-on-one chats with him about whatever is on their mind to schedule a time via this link; (2) Danny wants to host book-club-esque discussion groups with members—about books or movies or documentaries or whatever, ideally with some connection to NZN-ish themes—and he’d like suggestions along these lines (or just indications of interest in participating). You can leave this feedback in the comments section below.
Another thing you can leave there is the answer to this question: Would you join a NonZero Discord server if we launched one? (“What is a Discord server?” is an acceptable answer—but only if you’re over 50 and can provide proof of age.)
—RW
Eurovision, the annual song contest that is one of the world’s most watched non-sporting events, is facing a a crisis. Last week Spain—which has competed in every contest since 1961—became the fifth country (after Ireland, Slovenia, Iceland, and the Netherlands) to say it will not take part in the 2026 contest if Israel participates, marking the first time in Eurovision history that multiple nations have threatened a boycott.
Meanwhile, this week Russia, which was banned from Eurovision after its invasion of Ukraine in 2022, revived a Soviet-era alternative: Intervision, a song contest that during the Cold War was an Eastern Bloc event but this week hosted competitors from 22 countries, including China, India, and this year’s winner, Vietnam.
This competition between Eastern and Western blocs feels eerily familiar; it almost makes us wonder if the dim memory of Vladimir Putin singing Blueberry Hill in front of Kevin Costner and Sharon Stone was implanted—but video evidence from 2010 suggests otherwise. Those were the days.
Border control agents have for years collected DNA from US citizens in ways that may violate federal law, according to an analysis by Georgetown Law’s Center on Privacy and Technology. Government records show that some 2,000 citizens—and more than 2.6 million people overall—have had genetic material taken by the Department of Homeland Security and uploaded to the FBI’s criminal database since restrictions on mass sampling were loosened in 2020.
US Customs and Border Protection insists it acted within its authority. But the law permits collection only from citizens arrested, charged, or convicted of a crime, and, in some 500 instances, agents seem to have swabbed citizens who had merely been detained. Stevie Glaberson, research director of Georgetown’s privacy center, says the absence of adequate checks in the system “renders the program unconstitutional and violates the Fourth Amendment.”
Banners and graphics by Clark McGillis.
How about for Danny’s book club “If Anyone Builds It, Everyone Dies?”
"Like, for example: a massively lethal global pandemic caused by a novel bioweapon made abroad with AI’s help"
The right already massively resents alarmism over covid, so good luck on this...
Geoffrey Hinton thinks AI will create tremendous unemployment, as if getting past work is bad (even assuming such massive unemployment will happen). But he's also against UBI and wants to preserve work for the sake of it allegedly providing people dignity. How much dignity a guy who'd rather play Halo with online friends gets working part-time at Grocery Outlet is an open question.