Latest

Deep Dive — Amodei's dinner bought access, not a rulebook

The Guardrails

Deep Dive — Amodei's dinner bought access, not a rulebook

Dario Amodei had been trying to get this meeting for months. On Sunday night he got it, at the president's residence, over dinner — two days after the same government upheld a designation that bars the US military from using his company's models. The scoreboard on Friday Nothing announced at the dinner lifted that designation. The District of Columbia Circuit had ruled 2-1 on Friday that the Defense Department "had ample support for its conclusion that the continued integration of Claude into

AI 101 — What is LLM-as-a-judge?

The Frontier

AI 101 — What is LLM-as-a-judge?

If you read a week of AI news, you'll notice almost nobody grades models by hand any more. When a lab says its new model is better at writing, summarising or handling customers, the number usually came from another AI. An LLM judge is a language model used to grade the work of another model — or of your own AI feature — when no piece of code can decide whether the answer was good. The industry term is LLM-as-a-judge, and the job is simple to state: read a question and an answer, then return a s

OpenAI backed a startup built to fight an AI-made pandemic

The Guardrails

OpenAI backed a startup built to fight an AI-made pandemic

A biosecurity startup raised $36 million to design antibodies against the pathogens AI might help build — and it was already quietly firing at the attack surface it is defending. Separately, DHS started using AI to recommend what to redact from the records you asked for, and an anonymous model took third place on the world's busiest token leaderboard with a fingerprint no one was supposed to see. Red Queen Bio, a biosecurity startup whose $15 million seed OpenAI led in November, has now raised

A 27B security model from Alibaba beats the frontier labs on CyberGym

The Frontier

A 27B security model from Alibaba beats the frontier labs on CyberGym

A 27-billion-parameter security model that Alibaba fine-tuned itself now sits above Google's and OpenAI's purpose-built cyber models on UC Berkeley's vulnerability benchmark. Separately, the platform vendors started selling the tooling to explain what agents actually do — which is the question the first story is really about. Alibaba Security's XekRung-1.5-27B-Preview took the No. 1 row on CyberGym's model leaderboard with an 88.9% success rate at reproducing real vulnerabilities — 34.39 points

Insilico's AI-designed drug enters Phase III — a first for the field

The Frontier

Insilico's AI-designed drug enters Phase III — a first for the field

A drug whose target was found by AI and whose molecule was designed by AI is now in a late-stage trial — and the registry, not the press release, is what makes it a milestone. The same day brought a machine-checked maths result from a team of agents, and a brain-implant round with an unusual lead investor. Insilico Medicine has begun dosing patients in a 320-person Phase III trial of rentosertib, an oral treatment for idiopathic pulmonary fibrosis — the first drug with both an AI-identified tar

Australia's real AI problem is decades-old government software

The Guardrails

Australia's real AI problem is decades-old government software

Australia spent the week treating an AI agent as the intruder. Sunday's warning is that the house never had a lock. The same weekend, Beijing treated American doomsday talk as a negotiating position — and OpenAI put its training on hold. Australia's former chief UN cyber negotiator says the vulnerabilities that let an OpenAI agent reach Medicare data sit in the country's own ageing computer systems, and federal cabinet meets Monday on the fallout. Johanna Weaver, who led Australia's cyber negot

Today in AI — September 27, 2026

The Guardrails

Today in AI — September 27, 2026

Sunday's roundup runs heavier on rules than on models: Brussels' enforcement gap, a state attorney general asking Congress for a law, dark-web resellers undercutting the labs' own pricing, and Beijing selling the token rather than the model. The engineering that did land was about spending fewer tokens, not more. Models & Research * Fireworks Research shipped Ember-1, a version of Kimi K3 trained to stop over-thinking rather than to think harder. The company says K3 spends the majority of it

The Vatican tells the UN: keep the decision to kill with a human

The Guardrails

The Vatican tells the UN: keep the decision to kill with a human

The Vatican spent the week making an argument it has no power to enforce: that the decision to use lethal force cannot be handed to a machine, and that the rules governing AI have to be written globally. Also today, a benchmark that puts a price on honesty in web extraction — and finds a paid scraper less honest than a cheap model. Cardinal Pietro Parolin told the UN General Assembly that "the decision to use lethal force cannot be delegated to opaque or automated processes, but must remain und

The Week in AI — September 21–27, 2026

The Week in AI

The Week in AI — September 21–27, 2026

The week every answer was procedural. OpenAI kept widening the list of organisations its agents may have touched, Washington put itself in front of the pre-release gate while handing liability to the developers, Beijing and Washington gave the technology a name and a phone line, and the commercial question that will decide the next year was settled on price rather than capability. The week's top 5 1. The agent-swarm ledger got longer, and the answer to liability turned out to be contract law

Lambda will build a 375 MW Oklahoma data center with no subsidies

The Arena

Lambda will build a 375 MW Oklahoma data center with no subsidies

Two stories today: a neocloud that went shopping for a data-center site and found one that refused to pay for it, and a robotic hand that buys dexterity with linear motion instead of joints. Lambda announced Sunday that it will build a data center at MidAmerica Industrial Park near Chouteau, Oklahoma — a project the industrial park says comes with no financial incentives at all. Tulsa World reports the facility is planned at 375 megawatts on a 329‑acre site; Lambda's own release does not give a

Zhipu is down 79% from its peak as the model price war bites

The Arena

Zhipu is down 79% from its peak as the model price war bites

The token price war finally has a scoreboard, and it is denominated in shareholder value rather than tokens. China's most-watched listed model lab is worth about a quarter of what it was in June. Zhipu's Hong Kong-listed shares (02513.HK) closed Friday at HK$633.50, about 79% below the HK$2,980 peak they reached on June 22, putting the company's market value back near HK$309 billion. Zhipu listed in January at HK$116.20 a share and ran roughly 25 times that into June; it has now given almost al

Beijing tells Alibaba and ByteDance to file Nvidia purchase plans

The Guardrails

Beijing tells Alibaba and ByteDance to file Nvidia purchase plans

For a year, China's approach to Nvidia has been subtraction — stop ordering, buy domestic. Sunday's report is Beijing trying addition instead, on a chip that was never on the banned list. China's Ministry of Industry and Information Technology has asked Alibaba, ByteDance and other domestic companies to report their plans to buy Nvidia's RTX Pro 5500, and told them the government intends to approve the purchases. The Information's Qianer Liu reported the approach on Sunday, citing two people fa

See all