Superintelligence news
British AI neocloud Nscale raises $3.36 billion in convertible notes before IPO
The financing, led by Third Point and backed by Nvidia, provides $2.36 billion now and a further $1 billion later, as the company prepares for a $35 billion NYSE listing.
Google pilots direct purchases from Flipkart via Gemini and AI Mode in India
In a limited trial, Indian shoppers can click a Buy button on select Flipkart listings inside Gemini and AI Mode, triggering a checkout without leaving the AI interface.
Crusoe drops $1.25 billion turbine deal with Boom Supersonic
The Denver AI data-center firm ended its $1.25 billion purchase of 29 Boom Supersonic 42-megawatt Superpower turbines, citing a mismatch with its evolving energy strategy.
AI-driven claim submissions add nearly $1 billion to US healthcare spending
A Blue Cross Blue Shield Association study finds hospital AI tools inflated costs by $942 million over two years, while industry leaders warn the technology could deepen insurer-hospital tensions.
Federal Appeals Court Allows Pentagon to Blacklist Anthropic Over Claude Feature Refusal
The D.C. Circuit upheld the Department of Defense’s authority to designate the AI start-up as a supply-chain risk, citing broader statutory language and national-security concerns.
Meta’s Muse AI Agent Surpasses Early ChatGPT Benchmarks and Targets Wearables
Meta’s personal AI assistant Muse is reportedly delivering higher early performance than ChatGPT, while the company eyes integration with smart glasses and a miniature Tamagotchi-style gadget.
Thoughtful Things launches Engram sampler that reshapes AI hallucinations into sound
The Kickstarter‑backed instrument uses tiny AI models to glitch and invent audio, offering an open‑source firmware and a price point starting at $675 for early supporters.
Meta’s Muse AI Agent Surpasses Competitors in Latest Model Race
Meta’s Muse outperforms early ChatGPT benchmarks and is slated for integration into smart glasses and a compact wearable, shifting focus from OpenAI and Anthropic releases.
The UN Now Has a Standing Science Panel on AI, and Two Countries Voted Against It
Forty experts, a three year term, and a 117 to 2 vote that put the United States on the record against the body it will now sit outside of.
An AI Agent Got Into a Services Australia Portal, and Canberra Is Rewriting Its Posture
The breach happened on 18 June. The agency learned on 10 September. The Prime Minister said it out loud on 24 September, from the United Nations.
South Africa Pulled Its National AI Policy Because the Citations Were Not Real
A Cabinet approved draft, gazetted for public comment, cited sources that do not exist. The minister withdrew it and ordered consequence management.
China Wrote Rules for AI That Pretends to Be a Person
Mandatory disclosure, forced human takeover on self harm signals, a minors mode, and a ban on simulating a family member to an elderly user.
Japan Passed an AI Law With No Fines in It, on Purpose
Act No. 53 of 2025 bans nothing, licenses nothing and penalises nobody. The strategy is to be the most AI friendly country and to fund it.
Germany Wants a Gigafactory and Got a Regulator First
Brussels opened a call for up to seven AI gigafactories with 30 billion euros behind it. Berlin published the law that decides who polices the output.
Washington Is Accelerating Data Centres. Twenty Seven States Are Pricing Them.
Federal permitting relief does not reach zoning, land use or utility regulation, and that is exactly where the states are legislating.
Four Frontier Labs Shipped in Four Days
Anthropic, Google, Meta and OpenAI each released a frontier model in the first week of September 2026, and two of them shipped restricted cyber variants alongside.