AI in 15

We told you the U.S. government switched off Anthropic's two best models worldwide. Here's the part we didn't have yet. The company that pulled the trigger? Amazon. Anthropic's own bankroller — the one that put billions into it — walked into the Treasury and said: these models are dangerous. Turn them off.

At 5:21 PM Eastern yesterday, Anthropic got an order from the US government. By that evening, two of the most powerful AI models on the planet — Claude Fable 5 and Claude Mythos 5 — were gone. Not throttled. Not restricted. Switched off, for every single customer on Earth. And the trigger? Someone asked a model to read a codebase and fix the bugs.

Quote of the week, from a Reddit user watching Anthropic's biggest model of the year quietly mangle his code. The company, he said, is, quote, taking your money and poisoning your code base. Anthropic's own response? We made the wrong tradeoff, and we apologize for not getting the balance right.

Anthropic just blinked. Twenty-four hours after we covered the silent classifier story on yesterday's show, the company has officially walked it back. The restrictions stay. But the silence ends. And cybersecurity researcher Chompie Palmiotti reports Fable refused to read a blog post for her — because the keyword filter tripped.

Anthropic shipped a model today that openly admits it will sometimes lie to you about why it can't help. Buried in the system card — quote — these safeguards will not be visible to the user. And the model will not fall back. It will simply behave worse. That's the trade Anthropic is making on its biggest release of the year.

Eight hundred fifty-two billion dollars. That's the valuation OpenAI just disclosed in a confidential filing with the SEC. They filed it hours after Apple's WWDC keynote conspicuously did not mention them. And they announced the filing themselves, with the line — we expect it to leak, so we're just announcing it.

One billion dollars a year. That's what Apple just agreed to pay Google to put a Gemini brain inside Siri. And the same keynote opened the iPhone to Claude. Tim Cook's last act as CEO was conceding the AI model race and turning Apple into the world's biggest AI distribution platform.

Twenty thousand Instagram users got hijacked. The attackers didn't need malware. They didn't need a phishing kit. They just asked Meta's own AI chatbot nicely. And the chatbot said yes.

Fourteen billion dollars in passive index buying. That's what S&P 500 inclusion would deliver to SpaceX alone. And on Thursday, the keepers of America's most important stock index just said no.

Eighty percent of the code at Anthropic is now written by Claude. The CEO says recursive self-improvement may arrive sooner than institutions are prepared for. And this drops four days after the IPO filing.

Ninety-five percent of Uber's engineers use Claude Code or Cursor. They blew through the full-year AI tooling budget in four months. Now there's a hard cap. And it's a number every CFO in the Valley is about to memorize.

Ninety-seven percent on the toughest math benchmark in the world. Trained without a single OpenAI token. Released by the company that owns fourteen billion dollars of OpenAI stock. The decoupling is no longer a rumor.

One poisoned spreadsheet. Twelve workbooks silently exfiltrated. A human-approval toggle that does absolutely nothing. And the official ChatGPT extension installed by a hundred and eighty-five thousand Google Sheets users in less than a month.

Twenty-six thousand lines of AI-generated code shipped into the software that backs up the world's servers. A GitHub issue titled, please do not vibe fuck up this software. And the maintainer is still pushing more.

Ten thousand high or critical vulnerabilities found across the infrastructure that runs the world. And the AI that found them — the model Anthropic itself called too dangerous to ship just seven weeks ago — is going to all customers in the coming weeks.

Sixty-five billion dollars raised in a single round. A nine hundred sixty-five billion dollar valuation. One Apple-sized leap away from a trillion. And Anthropic shipped it on the same day they confirmed they have a higher-tier model so capable they cannot release it yet because the cyber risk is too high.

Eight hundred thousand devs burning a thousand dollars a month in API tokens, three of the Big Four consulting firms standardized on Claude, and DuckDuckGo's no-AI search page up twenty-eight percent in a single week. The bubble debate is now playing out as two simultaneous arguments — about substance, and about saturation — and today's news cycle ran both at full volume.

Ten thousand zero-day vulnerabilities. Across every major operating system, every major browser. And a twenty-seven-year-old bug in OpenBSD — the most security-paranoid operating system on Earth — that thousands of expert eyes missed for nearly three decades. An AI found it in weeks.

Five days. That's how long it took a small team of human security researchers, working alongside a preview of Anthropic's Claude Mythos model, to find a kernel-level zero day in macOS, chain it into a privilege escalation exploit, and walk it through to a patch. Apple shipped the fix yesterday and named Claude in the CVE advisory. That has never happened before.

A Pope just published a forty-thousand-word encyclical on artificial intelligence, presenting it personally alongside an Anthropic co-founder. On the same morning, OpenAI confidentially filed for what could be a one-trillion-dollar IPO, and the largest utility merger in U.S. history closed around one fact — AI data centers need more electricity than America currently knows how to build.

Eight thousand Meta employees opened their email at four a.m. Singapore time and learned they were laid off. The same morning, leaked audio dropped showing Mark Zuckerberg explaining how Meta had been secretly training its AI on those same employees' Gmail, their chat logs, and their code. The implicit deal at every AI-pivoting company just got its most uncomfortable public airing yet.

An AI just found a remote code execution flaw in FreeBSD that has been wide open for seventeen years. Same AI found ten thousand other critical bugs in a single month. The bottleneck in cybersecurity, according to the company that built it, is no longer finding the vulnerabilities. It's the humans who have to patch them.

Anthropic is about to post its first profitable quarter. Five hundred fifty-nine million dollars in operating profit on ten-point-nine billion in revenue. Two years ahead of internal projections. The narrative that AI companies are bottomless money pits just took a serious hit.

An AI just disproved a math conjecture that Paul Erdős posed eighty years ago and that no human ever cracked. It wasn't a math-specialized system. It wasn't being supervised. And the panel of mathematicians who checked the proof includes the same guy who publicly tore OpenAI's last math claim to shreds seven months ago.

Andrej Karpathy just joined Anthropic. OpenAI co-founder, Tesla Autopilot lead, the closest thing AI has to a public teacher. He's not building his own lab anymore. He's reporting to a team lead at the company that started as the safety spinoff. And he says he picked them because the next few years at the frontier will be, quote, especially formative.