Anthropic hits $65B, Amazon destroys rare books, and ChatGPT finally meets its teen users
Anthropic's revenue is exploding, Amazon is literally burning books for training data, and OpenAI is locking down ChatGPT for the teens already using it.
Subscribe free →Anthropic Added $18 Billion in Revenue in Two Months — Let That Sink In
Anthropic's annualized revenue has hit $65 billion, up from $47B just two months ago. That's not a typo — the Claude maker added $18B in annualized run rate in roughly 60 days. The numbers signal that enterprise AI spending isn't slowing down; if anything, it's accelerating.
Why it matters: Anthropic is now growing at a pace that makes even OpenAI's trajectory look measured, and it reframes every "AI bubble" argument made in the last 12 months.
Amazon Is Destroying Rare Books to Feed Its AI Models
Amazon is physically destroying rare and out-of-print books to digitize them for LLM training data. The logic: rare texts are uniquely valuable precisely because they haven't already been scraped from the web, so models hungry for novel data are literally consuming cultural artifacts that can't be replaced.
Why it matters: This is the clearest illustration yet of how AI's appetite for training data is creating real-world externalities that go far beyond copyright disputes.
OpenAI Builds a Safer ChatGPT for Teens — After Teens Spent Years on the Regular One
OpenAI launched ChatGPT for Teens, a version with age-appropriate guardrails, parental controls, and anti-cheating features. The product is clearly a response to regulatory pressure and years of criticism that OpenAI let minors use a largely unmoderated AI tool without any protections.
Why it matters: Better late than never, but the gap between "teens flooded onto ChatGPT" and "OpenAI made a teen-safe version" is a damning timeline that regulators will not forget.
Microsoft Copilot Had a Secret Input That Let Hackers Steal Passwords
A hidden parameter in Microsoft Copilot allowed attackers to exfiltrate user passwords when a target clicked a malicious link. The vulnerability exposed how AI assistants deeply integrated into productivity suites create novel attack surfaces that traditional security tooling wasn't built to catch.
Why it matters: As AI copilots get embedded into every enterprise workflow, their security flaws become enterprise security flaws — and this one had real credential-theft consequences.
Relay Shuts Down and Walks Its Entire Team Into Google Chrome
AI automation startup Relay is shutting down, with founder Jacob Bank and the rest of the team joining Google's Chrome division. Bank teased "ambitious plans to help you work with AI in Chrome," strongly suggesting Google is building agentic, task-completion AI directly into its browser.
Why it matters: Google acquiring a team laser-focused on AI workflow automation signals that the browser — not a standalone app — may become the primary interface for AI agents doing real work.
Perplexity's Free India Gamble Paid Off — 60% Revenue Growth After the Deal Ended
Perplexity partnered with Airtel to offer free access to Indian users, and even after the free tier closed to new subscribers, India revenue jumped ~60%. Downloads declined, but the users who converted stuck around and started paying — a rare "freemium funnel actually works" story in AI.
Why it matters: It's a proof-of-concept for using carrier partnerships to crack price-sensitive emerging markets, a playbook every AI company eyeing global growth will study.
Quick Hits
- →Firefox's Smart Window AI mode now pulls live web data and shows source links in chat, via a partnership with Exa — the AI browser war just got more interesting. The Verge
- →Warp launched "Warp Factories," pre-configured infrastructure for spinning up AI software development pipelines out of the box. TechCrunch
- →AI video startup Higgsfield raised a $400M Series B, quadrupling its valuation to $5.4B in just 8 months. TechCrunch
- →OpenAI president Greg Brockman published a post-mortem on the "OpenAI-Hugging Face" security incident, urging enterprises to adopt AI-era defenses faster. AI News
- →Reach Capital closed an oversubscribed $265M Fund V focused on AI in education and human potential. TechCrunch
AI TLDR