Anthropic raised $65 billion at a near-trillion-dollar valuation, dropped Opus 4.8, and revealed that its Mythos model has already found over 10,000 critical software vulnerabilities. Sam Altman, meanwhile, admitted he was "pretty wrong" about AI taking jobs — and the Pope published an encyclical on the subject.
The number that defined the week: Anthropic closed a $65 billion Series H on Thursday, valuing the company at $965 billion post-money and eclipsing OpenAI's $852 billion for the first time. The round was led by Sequoia, Dragoneer, Altimeter, and Greenoaks, with co-leads from Capital Group, Coatue, D1, GIC, and ICONIQ. Strategic infrastructure partners — Samsung, SK Hynix, and Micron — also came in, a signal that memory and chip suppliers want a seat at the table. A portion of the round, roughly $15 billion, consists of previously committed hyperscaler capital including Amazon's $5 billion from April. Anthropic said its run-rate revenue crossed $47 billion earlier this month. The company is now the world's most valuable private AI startup, and the IPO conversation has moved from "if" to "when."
Alongside the raise, Anthropic released Claude Opus 4.8 — described as a "modest but tangible" improvement over 4.7, with gains in agentic coding, reasoning, financial analysis, and what Anthropic calls honesty about its own progress. Pricing stays flat at $5/$25 per million tokens. The more notable product launch was dynamic workflows, which let Claude orchestrate tens to hundreds of subagents in the background, along with a new effort control panel that lets users decide how much compute Claude devotes to a response. The model shipped to Pro, Max, Team, and Enterprise users and is available across AWS, Google Cloud, and Microsoft Foundry. Anthropic also opened offices in Milan — its sixth European city — and Seoul, where Claude usage runs 3.5 times higher than population would predict. The company said it plans to triple its international workforce.
The security story deepened. Project Glasswing, Anthropic's restricted-access initiative using the Mythos model, published its first update: in roughly 30 days, Mythos and about 50 partner organisations identified more than 10,000 high- or critical-severity vulnerabilities in systemically important software. Cloudflare alone found 2,000 bugs, 400 of them high- or critical-severity, with a false positive rate better than human testers. Across 1,000-plus open-source projects, Mythos flagged 23,019 issues, 6,202 high- or critical-severity, with more than 90% validated as true positives. Notable finds included a 27-year-old remote crash vulnerability in OpenBSD, a 16-year-old flaw in FFmpeg, and a wolfSSL certificate-forging exploit scored at CVSS 9.1. The programme is running at a scale that will reshape how the industry thinks about defensive scanning — though the dual-use question Mythos raised in April hasn't gone away.
Microsoft made the week's most surprising admission: it's pulling internal Claude Code licences by June 30 after the pilot burned through its entire annual AI budget in months. Token bills for individual engineers were hitting $500 to $2,000 per month, and flat seat licensing had obscured the true consumption cost. Microsoft is steering developers back to GitHub Copilot. It wasn't alone — Uber disclosed that it burned through its entire 2026 AI tool budget by April, with nearly 84% of its engineers on Claude Code by March. The episode is the first loud signal that agentic coding tools, for all their productivity gains, carry an economic problem that usage-based pricing exposes immediately at enterprise scale. For Anthropic, it's a strange kind of compliment — the tool is so good engineers can't stop using it, which is exactly why it's being cut.
Sam Altman offered his own recalibration. In a conversation with Commonwealth Bank of Australia's CEO, Altman said he was "pretty wrong" about AI's impact on employment, admitting he'd expected more entry-level white-collar job losses by now than have actually occurred. "I'm delighted to be wrong about this," he said. He'd even tried having AI answer his Slack and email — and stopped, realising how much he valued genuine human interaction. Fortune noted that both Altman and Anthropic CEO Dario Amodei are walking back earlier apocalypse-adjacent predictions as both companies approach potential IPOs — a shift in tone that may be strategic as much as sincere. And on Sunday, Pope Leo XIV released Magnifica Humanitas, his first encyclical, a 42,000-word document on human dignity in the AI age. Published on the 135th anniversary of Rerum Novarum, it argues that technology is "never neutral" and calls for governance frameworks that protect labour, truth, and the common good. Anthropic co-founder Chris Olah attended the Vatican presentation. Whatever else this week was, it was a week where the biggest questions about AI — who funds it, who controls it, who it displaces, what it finds, and what it means — all landed at the same time.