POPE LEO XIV ISSUES FIRST EVER PAPAL DOCUMENT ON AI — BRINGS ANTHROPIC CO-FOUNDER TO THE VATICAN AND TELLS THE WORLD TO SLOW THE HELL DOWN

SOURCE: TIME

So here is something you did not have on your bingo card for 2026: the Catholic Church just dropped its first ever major papal document on artificial intelligence, and they brought the co-founder of Anthropic to the Vatican to help announce it.

Pope Leo XIV, who has been Pope for all of about a month at this point, published his first encyclical on May 25th. It is titled Magnifica Humanitas, which translates roughly to “the magnificence of humanity,” and the entire document is about protecting human dignity in the age of artificial intelligence. The Pope is worried. And when the Catholic Church is worried enough to issue an official teaching document about something, that is generally a sign that something has gotten quite serious quite fast.

What makes this genuinely interesting, and not just another famous institution saying “we too have thoughts on AI,” is that Chris Olah, the Anthropic co-founder who has spent years doing some of the most serious interpretability research in the field, was invited to stand at the Vatican and speak alongside the cardinals and theologians at the official presentation. That is an unusual combination of people to have in the same room. Olah is the guy who publishes dense technical papers about how neural networks actually work internally. The Pope writes encyclicals. And yet there they were together, because both of them are, in their very different ways, trying to understand something genuinely terrifying about the technology we are all building.

The encyclical itself calls on governments, corporations, and individuals to slow down. That is the core ask. It argues that the pace of technological development is outrunning humanity’s ability to understand what it is building, and that AI systems are beginning to affect human cognition, autonomy, relationships, and the very structures of power in ways that most people have not yet processed. The document is not anti-technology in a blanket sense. It is not the Catholic Church saying AI is evil and we should unplug the servers. It is more nuanced than that. It is saying that if a technology is powerful enough to reshape how human beings think, work, and relate to each other, then there is a moral obligation to govern it deliberately and with genuine concern for people who are not powerful shareholders in AI companies.

There is something genuinely striking about the timing here. This encyclical lands in the same week that Anthropic is on the verge of closing a funding round that will value the company at nearly a trillion dollars. The Vatican is asking the world to slow down and reflect. The investors are lining up to pour thirty billion dollars into the company that Chris Olah helped co-found. Both of those things are real at the same time, and the tension between them is not easy to resolve with a press release.

Olah’s presence at the presentation was framed by Anthropic as part of their effort to “widen the conversation” on AI’s important questions. That is corporate language, but the intent seems genuine coming from someone who has actually spent serious time thinking about what AI systems are doing internally. Whether any of the world’s governments will take the Pope’s advice and actually pump the brakes is another question entirely. The historical record on the world following papal guidance about technological development is mixed at best. But the fact that the most visible moral authority on earth just issued its first major statement on artificial intelligence tells you something important about where we are in this story. We have moved well past the point where this is just a tech conversation.

Keywords: Pope Leo XIV encyclical, Magnifica Humanitas, AI ethics, Chris Olah Anthropic, Catholic Church artificial intelligence

ANTHROPIC ABOUT TO CLOSE $30 BILLION ROUND AT $900 BILLION VALUATION — THREE MONTHS AGO THEY WERE WORTH HALF THAT

SOURCE: BLOOMBERG

Just three months ago, Anthropic closed a funding round at a $380 billion valuation. You read that correctly. Three months later they are about to close another one, this time at over $900 billion. The company has more than doubled its valuation since February, and the round is not even finalized yet.

Let that number actually land for a second. Anthropic does not make a consumer product most people have heard of. It makes Claude, which is genuinely excellent and widely used in enterprise software, but it is not a household name in the way ChatGPT is. And yet investors are about to hand over thirty billion dollars, or possibly more, at a valuation that puts the company ahead of OpenAI for the first time. OpenAI did its big round in March at $852 billion. If Anthropic closes above $900 billion, it becomes the most valuable private AI company on earth. That happened in about eight weeks.

What is driving this? The honest answer is a combination of real performance and market mania, and it is genuinely hard to tell where one ends and the other begins. On the real performance side: Anthropic reportedly hit a $30 billion annual revenue run rate, which is an extraordinary number for a company that barely had a product three years ago. The company achieved something like 80x revenue growth over the past period, which is the kind of growth you do not normally see outside of a pharmaceutical company that just got approval for a blockbuster drug. Enterprise customers are paying serious money for Claude’s API, coding capabilities, and reasoning features, and those customers are not leaving.

On the market mania side: venture capital firms and institutional investors are absolutely desperate to own a piece of frontier AI, and there are only two or three companies that genuinely qualify. Sequoia, Dragoneer, Altimeter, and Greenoaks are reportedly each putting in around two billion dollars to co-lead this round. These are not naive investors. They understand valuation math. But when there is a limited number of companies that can credibly claim to be building the most powerful AI systems in the world, and when those systems are generating tens of billions in revenue, the price goes up regardless of whether traditional valuation frameworks make any sense.

The wildcard here is the ongoing dispute with the U.S. Department of Defense, which blacklisted Anthropic earlier this year over its refusal to remove guardrails on weapons applications. That fight is currently in federal court. If Anthropic loses, or if the dispute drags on long enough to affect government contracts, that is a material risk to the business that investors are apparently willing to price as manageable. Anthropic is also reportedly on track for its first ever quarterly operating profit in Q2 of this year.

A company that is simultaneously growing at 80x, heading toward profitability, and commanding a nearly trillion dollar valuation is either the most important company in the world right now or one of the most extraordinary speculative bubbles in history. Possibly both. We will find out which one it is eventually, just probably not before the round closes.

Keywords: Anthropic funding round, Anthropic valuation, AI investment 2026, Sequoia Anthropic, Claude AI enterprise

OPENAI’S MOST FAMOUS BRAIN WALKED OUT THE DOOR AND STRAIGHT INTO ANTHROPIC — KARPATHY’S MOVE TELLS YOU WHO THE TALENT THINKS IS GOING TO WIN

SOURCE: TECHCRUNCH

If you have spent any time in the AI world and you know one name besides Sam Altman and Jensen Huang, it is probably Andrej Karpathy. He was one of the eleven original people who co-founded OpenAI back in 2015, then left to run Tesla’s Autopilot division, then came back to OpenAI, then left again to work independently, during which time he built a YouTube following of several million people who appreciate his ability to explain how large language models actually work. He is also the person who coined the term “vibe coding,” which you have heard approximately four thousand times this year.

On May 19th, Karpathy announced he was joining Anthropic. Not OpenAI, not Google DeepMind, not Mistral or any of the newer European labs. Anthropic. He will be working on pretraining, which is the most expensive and technically demanding part of building a frontier AI model. Pretraining is where you take an enormous amount of compute, an enormous amount of data, and run the training job that gives the model its core knowledge and capabilities. Everything else, the fine-tuning, the RLHF, the product packaging, is downstream of what happens in pretraining. When Anthropic hired Karpathy to work on pretraining and to build a team specifically focused on using Claude to accelerate pretraining research itself, they were not hiring for a communications role. They were hiring for the actual core of what AI labs do.

His framing when he announced the move is worth reading carefully. He wrote that “the next few years at the frontier of LLMs will be especially formative.” That is Karpathy telling you that he thinks the most important action is going to happen at the frontier level, not at the fine-tuning and deployment layer that most of the industry is currently focused on. He is making a bet with his career that whoever figures out how to push the frontier further and faster in the next few years is going to win the whole thing, and he has decided Anthropic is the right place to do that work.

For Anthropic, this is a coup that goes beyond the resume line. Karpathy is one of the most respected technical figures in the entire field. The fact that he chose Anthropic over OpenAI, where he has deep history and presumably deep relationships with a lot of the people there, sends a signal about how people inside AI actually perceive the two companies right now. OpenAI has more name recognition, more consumer reach, and a more complicated organizational history. Anthropic has a reputation for rigorous research culture, a clearer safety orientation, and now a nearly trillion dollar valuation that suggests the market agrees.

For OpenAI, losing Karpathy to a direct competitor after already losing him once before is the kind of thing that gets written into footnotes of industry histories. It is one person. But smart people with options choose their next move based on where they think the real intellectual work is going to happen and where they think their contribution is going to matter most. Karpathy looked at both options and picked Anthropic. That is the story.

Keywords: Andrej Karpathy Anthropic, OpenAI talent, AI pretraining, vibe coding, frontier AI models

ANTHROPIC DRAGS THE PENTAGON TO COURT OVER AI WEAPONS GUARDRAILS — THE MOST IMPORTANT AI LAWSUIT IN AMERICA AND ALMOST NOBODY IS PAYING ATTENTION

SOURCE: CNBC

You might have missed this one because it landed the same week as the Karpathy announcement and the Vatican news, but Anthropic opened its lawsuit against the Department of Defense in a Washington D.C. federal court on May 19th, and this case is worth understanding because it is not really about Anthropic. It is about whether AI companies can be legally compelled to strip safety guardrails from their systems as a condition of doing business with the U.S. government.

Here is the short version of how we got here. Earlier this year, Defense Secretary Pete Hegseth declared Anthropic a supply-chain risk, effectively blacklisting the company from Pentagon contracts. The underlying reason was that Anthropic refused to remove two specific restrictions from Claude: the model will not help design fully autonomous lethal weapons systems without human oversight, and it will not assist with mass surveillance of American citizens. The Pentagon wanted both of those restrictions removed. Anthropic said no. Hegseth declared them a national security risk and cut them off.

This was not a small decision. Anthropic had become deeply embedded in Pentagon operations. The military was using Claude for intelligence analysis, logistics, planning, and a variety of other tasks. When the blacklist landed, defense tech companies that had built products on top of Claude suddenly had a serious problem. Several dropped Claude and scrambled to find alternatives. The Pentagon began testing OpenAI, Google, and other models to find replacements. President Trump ordered the government to stop using Anthropic while giving the military six months to phase it out completely. Then Anthropic sued. And now a federal appeals court is actually hearing the case.

The legal argument Anthropic is making is essentially that the government cannot blacklist a company for refusing to do something that would make its AI more dangerous to civilians and to democratic oversight. The government’s argument is presumably that national security requirements are national security requirements, and private companies do not get to set the terms under which they operate in the defense space.

What is strange and significant about this case is that it could establish real legal precedent for how AI companies and governments negotiate about what AI systems can be used for. If the government wins, the implication is that a sufficiently motivated government agency can effectively force an AI company to remove safety restrictions by threatening to withhold contracts. If Anthropic wins, it establishes that refusing to remove safeguards against autonomous weapons use does not constitute a legitimate national security risk.

Every major AI company in America is watching this case closely, even if they are not saying so publicly. They all have things they say their models will not do, and they all have significant government contracts or ambitions to obtain them. The outcome of this lawsuit will establish the rules of that relationship for a long time to come. It is probably the most consequential AI legal case currently in front of any court in the United States, and it is getting a fraction of the attention it deserves.

Keywords: Anthropic Pentagon lawsuit, DOD AI blacklist, AI weapons guardrails, Claude AI military, AI legal precedent