"Useful Idiot": Why Peter Thiel Is Attacking the Pope Over Artificial Intelligence

Last Thursday, here in Berlin, the publishing group Axel Springer handed Peter Thiel its annual award. Outside, a few hundred people held signs reading "Thiel shall not pass". Inside, on a podcast recorded with chief executive Mathias Döpfner and released the same day, the co-founder of PayPal and Palantir explained why Pope Leo XIV is a useful idiot for the Chinese Communist Party: "The Communist Party of China is not going to listen to the pope. There's a chance people in the U.S. will. And so in effect, the Pope's encyclical was working for the Communist party." He pauses, allows that it may be too harsh, and says it anyway. It is his third attack in a year: first the lectures on the Antichrist, then Aspen in June, now Berlin. In the same interview he also found time to tell the Germans, his compatriots by birth, that they are afraid of success and that nearly all their rich inherited their money.

It is worth reading what the pope actually wrote, because Thiel does not dispute a single line of it. In the encyclical Magnifica Humanitas, published in May, paragraph 198 says: "it is not permissible to entrust lethal or otherwise irreversible decisions to artificial systems. No algorithm can make war morally acceptable." It goes on to say that AI in war turns defence into threat prediction, "reducing victims to data". Paragraph 5 asks the question that comes before any rule: who holds this power today, and how do they use it. Thiel has every reason not to want it asked.

A week before the interview, Bloomberg reconstructed the chain that on 28 February sent two US missiles into an elementary school in Minab, Iran: more than 150 dead, at least 123 of them children. The school, still filed as a Revolutionary Guards base, came out as a recommended day-one target from Maven, the Palantir system that the Pentagon, styled since September 2025 as the Department of War as well, uses to build its target lists. Central Command's civilian-protection team had gone from ten people to one. Palantir's answer was that it is not "responsible for the underlying data". Nobody in this story has clean hands: according to CSIS, Maven's language capabilities are supplied by Claude, Anthropic's model. Paragraph 198 reads as if it were written for that morning in Minab. I offer my reading as a hypothesis, and it is the simplest one available: Thiel is not attacking the pope over theology. The encyclical asks for limits on exactly the market his company lives on.

He has never been shy about what he thinks. In 2009 he wrote for the Cato Institute, a libertarian think tank: "I no longer believe that freedom and democracy are compatible." Palantir grew on public money from the start, beginning with the CIA's venture fund. In 2025 US Immigration and Customs Enforcement paid it $30 million for ImmigrationOS, a system for tracking immigrants marked for removal. In his lectures the Antichrist arrives in the shape of a regulator, someone who talks about safety and global coordination. A man who sells governments the software for choosing whom to bomb and whom to deport has found a theology in which the enemy of mankind is whoever asks him to stop. A convenient theology, one has to admit.

On the same 24 September, Axios revealed a memo circulating in the White House, written by a Trump political adviser with no official post, that describes effective altruism as a cult-like movement and names Dario Amodei, chief executive of Anthropic, as its face. In the days before, the voices had been different but in tune. David Sacks, co-chair of a White House advisory council on science and technology, had spoken of a "Doomer Industrial Complex". Emil Michael, a senior Pentagon official, of an ideology that is "pretty dystopian". Trump, of the whole risk as a "hoax". The pope, the head of a lab, the scientists: whoever asks for a limit is disqualified, each time under a different label, geopolitical, sectarian or theological. Coordination is not proven, and I do not present it as fact. The form is always the same: China will not stop, so whoever asks us to stop is working for China. It is the argument of the race without a finish line, carried all the way to the altar.

Of course the argument has a true core: a rule that binds only those who obey it shifts the advantage to those who ignore it. That is the problem of every arms treaty. The pope, however, does not write laws, and paragraph 198 stops no lab. What it asks is that the decision to kill stays with a person. Thiel's reply is to assign the pope to a geopolitical camp, without an argument.

The next wall stands on the other side of the equator. On 23 September, eleven days before Brazil's presidential election on 4 October, Meta suspended the advertising account of Lula's campaign on Facebook. It reinstated it the same day and did not tell Folha de S.Paulo why. It may have been a mistake. A mistake like that still needs an explanation, all the more from a company that in August, in an Amnesty International test, approved eight election-disinformation ads out of fifteen. Meta, for that matter, had announced which side it was on. In January 2025, when he ended fact-checking, Mark Zuckerberg promised to work "with President Trump to push back on governments around the world". He accused Latin American countries of having "secret courts that can order companies to quietly take things down", and Europe, meaning us, of "institutionalizing censorship". Six months later Trump imposed 50 per cent tariffs on Brazil, justifying them in part with Brazilian court orders against US platforms. Lula rejected that pressure in public. From here the hypothesis is hard to dismiss: an administration that uses tariffs to shield the platforms, and a platform that cannot explain why, for a few hours, it silenced that administration's main opponent in South America, days before a vote. From over here, you've forgotten what "conflict of interest" means.

This class of entrepreneurs tells itself it is libertarian, impatient with the state and its limits. The accounts tell another story. Palantir lives on public contracts, and the ceiling on the Maven contract was raised above $1 billion in 2025. Meta, in Zuckerberg's own words, can stand up to foreign governments only "with the support of the US government". Thiel put $15 million into JD Vance's 2022 Senate campaign. Vance, now vice president, told European governments in Paris in February 2025 that the future of AI "is not going to be won by hand-wringing about safety". Up close, the founders' genius is made of federal contracts, tariffs used as a shield and a vice president raised in-house. They are libertarians as long as the state is paying.

Which leaves the question of the advantage. In the same interview with Döpfner, Thiel admitted that the growing hostility to AI in the United States "is mainly what we're doing to ourselves". He is right, in a sense he may not intend. The walls knocked down this week were also guarantees for the people building the machines, the reason anyone agrees to use them:

- a moral voice speaking from outside the market;

- the rule that the decision to kill stays human;

- the neutrality of a platform days before a vote.

People who knock down the walls of their own house usually expect not to live in it.

The pope is in France this week, Brazil votes on 4 October, and the Pentagon's report on Minab, according to Bloomberg all but finished for months, has still not been released. Three things to watch, to learn whether this scorched earth is a calculation or a reflex.