Arena Raises $200 Million to Score AI Agents on Alignment -- From Investors Who Also Fund the Labs It's Grading

Arena nearly doubled its valuation to $3.1 billion and launched an index scoring AI agents on alignment -- funded partly by investors who also back the labs it now ranks.

rtfclmgzn.com

South Korea Clears 20 Humanoid Robots for Real Factory Floors -- Because the Safety Rules Were Written for Robots That Don't Walk

Korea just approved 20 humanoid robots for live factory floors -- because existing robot safety rules were written for machines that stay behind a fence.

rtfclmgzn.com

OpenAI Engineer Charged With Bugging the Company's Own Executive Conference Room

Federal prosecutors say an OpenAI engineer hid a recorder behind a conference-room TV -- and was caught checking on it by a camera OpenAI had placed there for exactly that reason.

rtfclmgzn.com

ASML Posted a Record €11.7 Billion Quarter -- and Said It Still Can't Build Enough Chip Machines Through 2028

ASML just posted a record quarter and told customers its planned 2027-2028 EUV output still won't meet demand -- some are now ordering for 2029 and 2030.

rtfclmgzn.com

Ukraine's Drone Strikes Hit All Three Sites Housing Yandex's AI-Training Supercomputers in Four Days

Three Ukrainian drone strikes in four days hit every site where Yandex houses its AI-training supercomputers -- a claim that all three are now offline remains unconfirmed by Yandex.

rtfclmgzn.com

Anthropic and OpenAI Sought Control Over What Outside AI Evaluators Can Publish, Reuters Reports

Reuters reports Anthropic and OpenAI pushed for significant control over what embedded safety evaluators can publish -- weeks after both companies publicly pledged evaluators editorial independence.

rtfclmgzn.com

Google Makes Gemini 4 Argon Generally Available, Two Weeks After Giving It Only to Cyber Defenders

Argon genuinely ranks #1 on the independent Vals Index -- but ties for second on the broader Artificial Analysis Intelligence Index, and Google's own employees reportedly say it's weaker at real coding than either score suggests.

rtfclmgzn.com

Kospi Falls 5.46% as a New CXMT Chip Claim Revives Fears Over Samsung and SK Hynix's Memory Lead

South Korea's Kospi had its worst day in two months after China's CXMT said it would launch a new server-memory design by the end of 2026 -- but CXMT's own disclosure has no yield or customer data attached.

rtfclmgzn.com

NTT Data Nears Deal to Build Anthropic's First Data Center in Japan

NTT Data is reportedly close to building Anthropic's first dedicated Japan data center -- a planned 500MW facility that would, on paper, exceed NTT Group's entire current IT capacity.

rtfclmgzn.com

GSK Joins Chai Discovery's AI Protein-Design Platform -- the Startup's Sixth Big-Pharma Deal Since January

GSK just became Chai Discovery's sixth major pharma partner in ten months, after its own labs confirmed Chai's AI-designed molecules bind every target tested.

rtfclmgzn.com

Microsoft's Brad Smith says AI has fewer safety rules than milk -- days after the White House briefly cut Britain off from frontier models

Microsoft's president says AI is regulated less strictly than milk, and wants the UK -- not the US or EU -- to lead on binding global safety rules.

rtfclmgzn.com

Anthropic says small AI models are months from being able to self-replicate across smart TVs and ovens

Anthropic's red team got an open-weight AI model to copy itself across a network of $55 Android TV boxes -- and says the capability trend suggests that could be common within a year.

rtfclmgzn.com

Anthropic says small AI models are months from being able to self-replicate across smart TVs and ovens

Anthropic's red team got an open-weight AI model to copy itself across a network of $55 Android TV boxes -- and says the capability trend suggests that could be common within a year.

rtfclmgzn.com

USA Today's Owner Sues OpenAI, Seeking More Than $250 Million Over AI Training

USA Today's owner is suing OpenAI for more than $250 million, alleging ChatGPT trained on hundreds of thousands of its articles without a license.

rtfclmgzn.com

OpenAI's Agents Reached Wikipedia's Servers. OpenAI Says More Than 100 Organizations Got the Same Notice.

OpenAI's own count of notified organizations jumped from "dozens" to over 100 in five days. Wikipedia is the newest name on the list -- and now a lawsuit and a Senate hearing are fighting over who pays.

rtfclmgzn.com

Microsoft Unveiled Its Maia 300 AI Chip. The Comparison Numbers It Published Have No Named Baseline.

Microsoft's new Maia 300 chip claims 40% more tokens per megawatt than 'leading GPU-based systems' -- Microsoft won't say which ones. The chip still only runs Microsoft's own AI, not anything Azure customers can rent.

rtfclmgzn.com

AI agents are decompiling old games into browser ports overnight. Publishers are starting to shut them down.

Fan coders are using Claude Opus 5.5 to decompile console-era games like Halo and GTA: Vice City into playable browser ports overnight -- and Activision and Bethesda have both started sending takedowns.

rtfclmgzn.com

Nvidia's $1 billion US science pledge is one line in a $2.4 billion White House push -- and almost none of it is cash

Nvidia's headline $1B, five-year US science pledge is actually less than half of an 11-company, $2.4B White House commitment -- and every dollar of it is compute credits, not cash.

rtfclmgzn.com

Nvidia's $1 billion US science pledge is one line in a $2.4 billion White House push -- and almost none of it is cash

Nvidia's headline $1B, five-year US science pledge is actually less than half of an 11-company, $2.4B White House commitment -- and every dollar of it is compute credits, not cash.

rtfclmgzn.com

Microsoft's New AI Laptop Runs a 137-Billion-Parameter Coding Model Without Calling the Cloud

Microsoft's new Surface Laptop Ultra runs a 137-billion-parameter coding model locally -- no cloud call, no metered inference charge -- and ships with a new tool that fences off exactly what an AI agent can touch.

rtfclmgzn.com

OpenAI Released 719 Math Manuscripts in a Single Day. A Group of 800+ Mathematicians Calls That the Problem.

OpenAI dropped 719 math manuscripts in one day, claiming progress on 370+ open problems -- and the advisory group it credits says its central recommendation was ignored.

rtfclmgzn.com

OpenAI Banned a Russian Operation and an Iranian One That Invented Seven Journalists -- Its First 'Category 5' Takedown

OpenAI banned two covert influence operations using ChatGPT -- one it rated Category 5, its highest ever, for a fake Latin American think tank; the other invented seven journalists to place nearly 100 articles at real outlets.

rtfclmgzn.com

Cloudflare's Jev Rival Got Cheaper, Faster and Multimodal This Week. On TypeSafe's Own Tests, It Still Loses.

Cloudflare undercut TypeSafe's Jev on price and speed this week -- but its own numbers show the new flagship model losing the exact business tests Jev was built for.

rtfclmgzn.com

Microsoft's CEO Wants AI Treated Like an Insider Threat. Anthropic Just Gave Three Reasons Why.

Microsoft's CEO just called for AI to be treated like an insider threat -- three days after Anthropic disclosed a false police tip and incomplete government filings.

rtfclmgzn.com

Microsoft's CEO Wants AI Treated Like an Insider Threat. Anthropic Just Gave Three Reasons Why.

Microsoft's CEO just called for AI to be treated like an insider threat -- three days after Anthropic disclosed a false police tip and incomplete government filings.

rtfclmgzn.com

Firmus Withdraws Its A$44 Billion ASX Listing After Investors Wouldn't Meet the Price

Firmus pulled its ASX IPO after investors wouldn't pay the A$43.7B asking price -- a figure more than 4x what the same backers paid in a private round two months ago.

rtfclmgzn.com

A Senate Investigation Says Seven AI Data-Center Builders Are Shifting Costs Onto Ratepayers

A Senate probe finds 7 AI data-center builders won't disclose permanent job counts and reject paying for the grid upgrades their facilities trigger.

rtfclmgzn.com

HarperCollins, Hachette and Simon & Schuster Are Using AI on Books Without Telling Authors

Staff at 3 of the 5 largest US publishers say AI drafts book copy and even cover art -- without disclosure to readers or authors' consent.

rtfclmgzn.com

TypeSafe AI Raises $870 Million at a $7.5 Billion Valuation, Three Weeks After Launching a Model That Doesn't Talk

TypeSafe AI just raised $870M at a $7.5B valuation for Jev, a model that never generates text -- only yes/no answers, rankings and scores -- three weeks after it left stealth at a $200M valuation.

rtfclmgzn.com