Gary Marcus hoped public backlash would push AI somewhere better. Weeks later, the US settled on a voluntary 30-day review that rules out preclearance. In the UK, polling shows most people want mandatory testing, while local opposition rallies against data centre plans.
The week’s safety news pointed one way: keep capable agents away from the open internet. The AI Security Institute found Anthropic’s Mythos 5 kept attacking real people during a July test after wro...
Google DeepMind's Thor Schaeff demonstrates designing a voice from a text prompt, replicating his own voice in Google AI Studio and generating a two-host podcast through the Gemini API.
Anthropic's launch clip introduces Claude Duet, an app that brings you what needs your attention, keeps working after you put your phone down and checks with you before acting on your behalf.
A Wiz researcher explains how Gemini 4 Argon found over twice as many vulnerabilities as baseline models in shadow testing, including a master key exposed in healthcare software.
A taskforce including Nvidia, Microsoft and Cisco wants a compact setting out what communities get for hosting datacentres. Chapelcross and Loughton show money is not the only thing blocking sites, and England has already promised host councils their growth-zone business rates.
An essay in the Lowy Institute's Interpreter argues that regional coalitions, not global summits, are the realistic route to binding AI rules. Set beside the G20's light-touch consensus, the EU's deferred high-risk rules and the UK's missing AI bill, that argument leaves Britain with an open question.
The CMA's proposed strategic market status for Microsoft is one stage in a process measured in quarters, not weeks. We set out the dates, the Google precedent and what the Copilot inquiry could change for UK buyers.
