The current week of AI briefings, Monday to Sunday, in plain English. The weekly summary is on top; each day's full report is linked below. Updated every morning and again mid-afternoon on weekdays; a new week starts each Monday.
The week opens with Nvidia at the center of two very different stories. Nvidia's board opened an independent investigation after reporting that CEO Jensen Huang was the target of a foreign intelligence effort - and Huang admitted an affair in a memo to employees. The same weekend, Nvidia partnered with Korea's Rebellions and SK Telecom on sovereign AI infrastructure built around Korean-made chips.
In Washington, the small federal office that assesses frontier AI safety - leaderless and no longer publishing its research - has been asked to informally report to the White House, the Associated Press reports. And on Monday afternoon, Google widened its TPU chip lineup with two new systems - one built for massive context windows, its first that keeps data encrypted while in use - in what Bloomberg calls an escalation of its rivalry with Nvidia.
On Tuesday, Anthropic's first concrete IPO price - roughly 867 million shares at $70 to $80, raising up to $69.3 billion - reached investment banks and was withdrawn the same day as inaccurate, days before the planned roadshow. Meanwhile, eight countries and the major AI labs signed a Paris declaration committing to shutdown mechanisms, outside audits and risk disclosure for powerful models, and Meta released Muse Ember, a cheap small model built for the repetitive work agents do. In the afternoon the model race took over: Google opened Gemini 4 Carbon, its strongest model, to paying users, Anthropic shipped the faster, cheaper Fable 5.2 and Mythos 5.2 pair, and Nvidia's GR00T 2 showed robots copying skills demonstrated only by humans.