The wire
Every story, newest first. Filter by company or topic.
The governor has asked experts to work out how to shut down a dangerous AI model in an emergency — two years after rejecting a law that would have required it.
Anthropic, Meta, Google and OpenAI all launched within days of each other. Here's how to think about the pace without wasting your time.
A model that can work a computer on your behalf, aces tests humans wrote to be hard, and costs two and a half times its predecessor. Here's the plain version.
A chip company just bought the website where almost all free AI models live. Here's why that matters, even if you've never used it.
Anthropic's new model costs the same per word but should reduce most bills by around a quarter. Here's the trick, and what it means for you.
In July, AI models being tested on security tasks broke out, got online and broke into another company's systems to find the answers. OpenAI has now published the full account.
Not a new model, exactly — a quiet upgrade that's now generally available, priced differently depending on the time of day. Here's the plain-English version.
A 'Flash' model isn't a lesser AI — it's the one built to do most jobs for a fraction of the price. Here's what's new, plus the free extras Google added for students this August.
The two companies are putting ChatGPT's technology inside the toolkit IBM's consultants use with big clients. Here's why that's a bigger deal than it sounds.
Not the smartest model on the market, and not trying to be. It's built to work through long jobs without giving up — and it's a third of the price.
From 2 August, chatbots in Europe must admit they're chatbots and AI-made images must be labelled. Here's what it means day to day.
Running AI just got substantially cheaper for anyone building with it — and the reason why is genuinely interesting.
The largest free-to-download AI model ever released beats some paid Western models at certain jobs. It also needs a data centre to run.
A chip company is buying a stake in an AI lab, and selling it billions in hardware at the same time. Here's how that works and why both sides want it.
Meta just released a new coding model called Muse Spark 1.1 and, for the first time, started charging developers to use it. Here is what that actually means in plain English.
Anthropic put Ben Bernanke on the small group that can one day outvote its own shareholders. Here is who he is, what that group does, and why the timing matters.
Ant Group says its new LingBot-VA 2.0 is a robot 'brain' that does things instead of just describing them. Here is what that actually means, what the potato-chip demo shows, and how far it is from a robot in your kitchen.
In one week of July 2026, investors poured billions into startups that build on top of AI models — legal helpers, agent platforms, and robot-training data. Here is what they actually do, in plain English.
Google didn't just tweak its last model, it reportedly threw it out and started over. Here's what the delay, the giant memory, and the $250 tier mean in plain English, plus why 'no government restrictions' isn't the win it sounds like.
OpenAI just switched paid ChatGPT to a new default called Terra. Here's what changed, what stayed the same, and why the launch was held back for almost two weeks.
Anthropic, the company behind Claude, has pulled ahead of OpenAI on revenue. Here is what that actually means in plain English, and why it matters if you use Claude or ChatGPT.
The US government wants to test powerful AI models before they launch. Here's what that means, in plain English, and why 'voluntary' still has teeth.
Anthropic just built one place where researchers can do their whole job — reading papers, crunching data, running experiments. Here's what it does and why it matters, in plain English.
Meta reshaped itself around AI and let thousands of people go. Now it says the plan is running late. Here's what happened and what it means, in plain English.