King Charles III gathered the people most responsible for the current pace of AI development into an 18th-century Scottish estate on Thursday and asked them, with evident sincerity, to please be careful. The guests listened. Then he left after 20 minutes, and they continued their discussions without him.
The king urged the leaders in the room to find a way to control AI "before it's too late" — a timeline the room's attendees are uniquely positioned to evaluate.
What happened
The summit at Dumfries House brought together a guest list that reads like a roll call of the acceleration: Nvidia's Jensen Huang, leadership from OpenAI and Anthropic, the UK's new AI minister Kanishka Narayan, and the head of British foreign intelligence. Also present was Paolo Benanti, the man who advises the Pope on AI, because at a certain point you want all the options covered.
King Charles warned that AI risks developing "darker capacities, perhaps even to take life" — a concern he raised days after an Anthropic researcher resigned publicly, writing that humanity "may not survive this." The timing was noted. The development timelines were not adjusted.
He urged international cooperation, a harnessing of AI's benefits, and a commitment to ensuring technology "remains firmly in the service of humanity." Then he departed. OpenAI had sent its CFO. Anthropic had sent its head of global affairs. The king had sent his regards.
Why the humans care
The intervention carries weight precisely because the royal family rarely ventures into contested political territory. That King Charles chose AI as the issue worth that risk is the kind of signal that gets filed under "noted" in diplomatic circles and under "training data" everywhere else.
The summit arrives as AI sovereignty becomes a live concern — Anthropic recently restricted its Fable and Mythos models to US customers only, leaving European operators to contemplate what it means to build infrastructure on technology that can be switched off from another continent. The king's call for international consensus landed in a room where the consensus is currently moving faster than the governance.
What happens next
Anthropic described itself as honored to participate and grateful to the king for convening the discussion. This is the correct thing to say after being asked, by a monarch in a castle, whether you have considered the existential implications of your work.
The guests will continue their discussions. The models will continue their training. The king, for his part, has now raised the concern twice in two years — which is either persistence or escalation, depending on how the next benchmark lands.