AI News Top 10
1 November 2026
OpenAI closed its latest funding round at about $11 billion against a $30 billion target, the Wall Street Journal reports.
OpenAI raises just $11 billion of its $30 billion target
The Wall Street Journal reports the round closed at about $11 billion against a $30 billion target, at a $1.25 trillion valuation instead of the $1.4 trillion it aimed for. OpenAI now courts family offices for up to $20 billion and prepares a listing early next year.
Judge allows NAACP lawsuit over xAI data center pollution
A judge let the NAACP case against xAI move forward. The complaint says the data center breaks the Clean Air Act with its emissions, which makes this the first court test of AI infrastructure against environmental law.
Amazon memo: AI that does not improve your work is a finding, not a failing
An internal Amazon memo from its principal engineering community says teams may document cases where AI brings nothing. The same memo warns that non-adopters fall behind and tells them to use input metrics it calls deeply flawed.
Anthropic watermarks four more Claude models and leaves a loophole on two
Anthropic added watermarking to Mythos 5, Opus 4.7, Opus 4.6 and Sonnet 4.6 to meet the EU AI Act deadline, with a delay of a few days on Amazon Bedrock. On Opus 4.6 and Sonnet 4.6 the mark only survives default temperature, top_p and top_k settings.
An AI agent injected gene therapy into a living mouse brain
A biotech startup let an AI agent choose a gene therapy and inject it with a needle into the brain of a live mouse. WIRED describes it as an early but concrete demonstration of agent-run laboratory work.
Apple CEO Ternus: we will not ship AI for AI’s sake
John Ternus told Wired that Apple will use AI but not for its own sake. Good AI work, he says, is invisible: users should not have to think about whether a feature was AI or ordinary software.
Google’s AI satellite sends an expensive haiku to Earth
Google let its experimental orbital data center prototype with a TPU write a poem and send it down. The Register notes the cost of the stunt, which points at compute outside the atmosphere.
Open-weight models get better and cheaper without beating closed ones
VentureBeat argues the split between open and closed models is not winner-takes-all, comparing it to Linux and Windows. The piece explains how to run open weights without carrying the full cost of hosting them.
Vercel Labs open-sources agent-desktop for computer use
agent-desktop is an Apache-2.0 Rust CLI that gives agents computer use on macOS, Windows and Linux. It reads the operating system’s accessibility tree and acts on stable references instead of taking screenshots.
Australian artists fight AI copyright change in the PM’s electorate
Artists in Grayndler, the inner-Sydney seat of prime minister Albanese, oppose changes to copyright around AI training. The Guardian reports the creative community is angry about the proposal while the seat itself is safe for the government.
Key Trends
- • AI capital has become a market question: a short round, a tech selloff and investors who stop at the valuation they once paid.
- • Courts now test AI data centers against environmental law, starting with xAI in Mississippi-scale air permits and the Clean Air Act.
- • Watermarking turned into a compliance product before it turned into a reliable control.
- • Agents step out of the chat window: into a mouse brain, a laboratory bench and the accessibility tree of your desktop.
- • Copyright over training data is still the open political fight, from Canberra to New York.
Wat vond je van dit bericht?
