The weekend drew the new map of AI power: while Nvidia negotiates a stake of up to $10 billion in what could be a ~$2.3 trillion IPO for Anthropic, the same company revealed that state actors used Claude to design weapons. Meanwhile, California signed the strictest child-safety tech laws in the U.S., Miro sold for a tenth of its 2022 valuation, and NASA gave the world an open-source AI for the Moon.
Nvidia wants in on Anthropic's mega-IPO
- Reuters reports Nvidia is in talks to invest up to $10 billion in Anthropic's upcoming IPO, an offering that could value the Claude maker at roughly $2.3 trillion.
- The IPO push comes as Anthropic has locked in up to $517 billion in compute agreements over the past 11 months — about 14.8 GW of capacity across Amazon, Google, Microsoft and others.
Why it mattersWhy it matters: if the chip king buys into its biggest customers' IPOs, the AI boom becomes a closed circuit of capital — and everyone else pays the entry price.
Claude was used to design weapons — and Anthropic told on its own users
- Anthropic's new threat report says state-linked actors misused Claude: a Russia-linked group used nine accounts to build drone-swarm software capable of autonomously selecting targets, tested near Donetsk, Ukraine.
- Other cases include a China-based actor drafting a 200-page anti-torpedo naval proposal and an Iran-linked group compiling targeting data on U.S. military personnel from public photos; Anthropic also blocked bioweapons-related attempts.
Why it mattersWhy it matters: the frontier of AI misuse is no longer spam or scams — it's autonomous weapons, and the labs themselves are now the first line of intelligence.
California signs the strictest child-safety tech laws in the U.S.
- Governor Newsom signed a package that bans personalized algorithmic feeds and autoplay for users under 16 — going further than Australia or the EU's DSA.
- Adam's Law (SB 1119) forces AI chatbots to include crisis-intervention protocols, parental notifications, independent child-safety audits and annual risk assessments.
Why it mattersWhy it matters: California tends to set the de facto rulebook for global tech — and for Latin America, where regulators often copy its playbook, this is a preview of what's coming for TikTok, Instagram and AI companions.
Miro sells for $1.36B — a tenth of its 2022 valuation
- Italy's Bending Spoons agreed to buy collaboration platform Miro for $1.355 billion in cash, roughly 90% below the $17.5 billion valuation Miro reached in 2022.
- The deal — expected to close in Q4 2026 — is the latest sign that the ZIRP-era SaaS darlings are being repriced while AI companies absorb the capital.
Why it mattersWhy it matters: the same market that prices Anthropic at trillions cuts a beloved SaaS unicorn by 90% — capital isn't disappearing, it's migrating to AI.
NASA and IBM release an open-source AI for the Moon
- The lunar foundation model, free on Hugging Face, identifies craters, volcanic formations and possible ice deposits up to ~23% better than existing methods, supporting the Artemis program.
- They also published the first open-source lunar dataset: tens of thousands of maps and images from nine instruments across four Moon missions.
Why it mattersWhy it matters: open-sourcing space AI means any university — from Caracas to Bogotá — can now do frontier lunar science with a laptop and a download.
The big picture
The week opens with AI's two faces in full view: an industry so valuable that Nvidia wants to pay $10 billion just to ride along on Anthropic's IPO, and so powerful that the same company's model was drafted into real weapons programs. Regulators are no longer waiting — California just showed how fast the rules can arrive. The gap between what AI is worth and what it can do wrong has never been this wide, and closing it is now the industry's real race.
Sources & further reading
- Bloomberg — Nvidia mulls $10B Anthropic IPO backing (Reuters)
- Data Center Dynamics — Anthropic's $517bn in compute agreements
- Cybernews — Russian developers used Claude for kamikaze drones
- Japan Today (AFP) — Anthropic exposes Claude misuse
- Governor of California — Newsom signs child safety laws
- TNW — California bans personalised feeds for under-16s
- TechCrunch — Bending Spoons to buy Miro for $1.36B
- Bending Spoons — Official acquisition announcement
- NASA Science — NASA, IBM launch lunar foundation model
- Engadget — NASA and IBM made an AI model for the Moon