The 90-Minute Shock Heard Around the G7
The day's biggest story is also its most disruptive: the Trump administration abruptly ordered Anthropic to cut off access to its newest AI models — reported as Fable 5 and a sibling model — for all foreign nationals. The restriction is sweeping, reaching users inside the United States and even Anthropic's own employees. According to The New York Times, the company got roughly 90 minutes of warning, throwing the high-profile firm into confusion.
The timing could hardly be sharper. The move landed just as global AI leaders gathered for the G7 Summit, and it is already rippling through Washington's relationships with key allies. Reporting from CNBC, Bloomberg, KSAT and others paints a tense picture: G-7 partners are bristling, and world leaders — with the French president out front — are pressing the U.S. to share its most advanced AI while openly worrying that Washington could pull the plug at any moment. In short, allies want American AI, but not an American off switch.
That anxiety played out against an unusual backdrop. Some of the most powerful figures in the field turned up alongside President Trump at the summit, with OpenAI's Sam Altman and Anthropic's Dario Amodei breaking bread in the same room as heads of state. The chiefs of Anthropic and DeepMind used the moment to make a case for a U.S.-led AI coalition, arguing America should take the lead in writing the global rules. The export clampdown, however, complicates that pitch considerably.
A Marquee Defection
The other headline is a talent shake-up at the top. Noam Shazeer — a Google vice president of engineering and co-lead of its Gemini models — confirmed Wednesday that he is leaving for OpenAI, per Reuters. Losing a Gemini co-lead to its fiercest rival is a notable blow to Google and a coup for OpenAI, underscoring how fluid the loyalties remain among the industry's most sought-after engineers.
Products Keep Shipping
Even amid the geopolitics, the build cycle didn't pause. Anthropic is tightening the connection between two of its tools, updating Claude Design to work more smoothly with its popular coding agent, Claude Code, according to VentureBeat — stitching design and coding into a single flow. OpenAI, meanwhile, rolled out LifeSciBench, a benchmark meant to test how well frontier AI handles the messy, real work of life-science research rather than biology trivia.
AI is also burrowing deeper into the tools that build entertainment. Epic Games detailed Unreal Engine 6, folding generative AI — including Claude and Gemini — directly into the software behind many of the world's biggest games. And Nvidia's long-promised AI teammate finally went live: its ACE-powered "PUBG Ally" is now playable in the battle royale shooter, per Wccftech.
The Public Isn't Sold
For all the boardroom and summit-room momentum, the public mood is cool. A new Pew Research study, reported by TechCrunch, found just 16% of Americans believe AI will make society better. Wall Street may be betting big, but Main Street remains unconvinced — a gap worth watching as Washington, Silicon Valley and world leaders all scramble to shape what comes next.
A day, in other words, that captured the whole contradiction of this moment: AI woven ever more tightly into games, labs and software, even as governments fight over who controls it and the public wonders whether any of it is good news.