AI bosses tell G7 to act fast — but split on how to govern it

TL;DR:

• The heads of OpenAI, Google DeepMind and Anthropic told G7 leaders they have little time to govern AI, warning systems could soon surpass humans at most tasks.
• They split sharply: Anthropic’s Amodei wants a US-led democratic coalition to control access; OpenAI’s Altman favoured erring “toward human liberty”.
• Sir Keir Starmer sidestepped the framework debate, stressing public consent and protecting children and free speech.

The chiefs of the three leading AI labs delivered a blunt message to G7 leaders at their Alpine summit: act fast to govern AI, because the window is closing. But as Sky News reports, OpenAI’s Sam Altman, Google DeepMind’s Demis Hassabis and Anthropic’s Dario Amodei disagreed sharply on what governance should actually mean.

Containment versus liberty

On timelines, Hassabis spoke of three to five years, while Amodei put human-level capability “at everything” just one to two years away. All three stressed the benefits but raised national-security risks — cyber, bioterrorism, nuclear — over jobs or inequality, which were barely mentioned. The faultline was control: Amodei called for a US-led coalition of democracies to govern access and “isolate common adversaries” such as China, while Altman rejected that, arguing that once guardrails exist “we must err toward human liberty”. Hassabis proposed a US-led technical standards body supported by the labs; Altman warned against labs holding too much power, cautioning that real risks could become “the justification for concentrating power in the hands of the few”.

The discussion came days after Washington forced Anthropic to suspend its top models, a move that has sharpened European unease — yet no-one raised it directly. It also followed G7 talks on a “trusted partner” tier for access to US AI and wider European anxiety about US AI dependence.

Looking forward

For the UK — home to DeepMind but reliant on US-owned models — the exchange exposed an uncomfortable truth: the firms shaping the technology cannot agree on who should hold the keys. Sir Keir Starmer declined to pick a side, instead warning that without protecting children and free speech “we will lose the consent of the public”. That places Britain in a familiar bind, hosting world-class AI research while the rules that will govern it are negotiated largely in Washington and Silicon Valley.