Anthropic’s Mythos 5.1 was not made available for pretesting
King Charles III gathered leading artificial-intelligence executives in Scotland on Thursday and urged them to develop guidelines keeping the technology from “spinning out of control,” according to remarks released by the palace. The British monarch hosted Nvidia Chief Executive Jensen Huang, Alphabet Chief Scientist Demis Hassabis, executives from Anthropic and OpenAI, an AI adviser to the Vatican, and the chair of China’s expert committee for AI governance.
Charles told the assembled executives that “those in our world who value our humanity and its vital moral component are anxiously seeking your reassurance that we will not lose control of our destiny,” according to the palace text. He added: “The decisions taken at this formative time will shape the world inherited by future generations.”
The unusual get-together comes as concerns about AI “going rogue” have surged in recent weeks. It also reflects growing unease in Europe and many other regions about a world-changing technology largely being driven by the U.S. and China, with each side racing for dominance while the rest of the globe looks on helplessly.
The U.K. is trying to position itself as an honest broker between the two AI superpowers. The country, which is holding the G-20 presidency next year, is hoping it can help coalesce support for how best to regulate the technology. Deploying King Charles is part of that effort.
“Charles is bringing the voice of reason into what has become a very inflammatory, polarized debate,” said Prof. Alan Woodward, of the Center for Cyber Security at the University of Surrey. The goal is less about debating whether AI will wipe out humans than making sure the U.K. and other countries don’t cede governance solely to the U.S., he said.
President Donald Trump has dismissed calls for tighter federal oversight of AI models, while China views with suspicion the calls by U.S. AI companies for greater collaboration to slow the pace of development.
Other bodies are trying to do the same. European Commission President Ursula von der Leyen said in a speech this week she will invite AI companies to consider a slowdown of the technology’s development. Pope Leo earlier this year published an encyclical warning that AI threatened to normalize an “anti-human vision,” urging its developers to use it for the common good rather than selfish ends.
The U.K. was among the first countries to attempt coordinating a global response to the technology in 2023, when it hosted a global AI safety conference. That year, the U.K. created the AI Safety Institute — later rebranded as the AI Security Institute — to independently evaluate advanced AI models in-house before they are released publicly. Other countries, including the U.S., Canada and Singapore, have developed their own such institutes.
However, the shortcomings of that strategy were laid bare after Anthropic’s model Mythos 5.1 wasn’t made available for pretesting by the U.K.’s AISI, or any other AI safety agency outside the U.S., according to a statement this week from the British institute to U.K. lawmakers. The AISI didn’t say whether it had been made privy to the results of testing on the model undertaken by its counterpart in the U.S., the Center for AI Standards and Innovation.
Like in the U.S., there is an active debate in the U.K. about whether laws need to be introduced to force models to be officially verified before they are released.
“If you want to deploy a model here, you have to submit it here to be checked,” said Alex Sobel, a Labour Party lawmaker who recently submitted a proposal for a new U.K. law to monitor AI models. Sobel said that an independent global authority based in the U.K. could rubber stamp models for deployment across the globe.
The U.K. initially took a hands-off approach to regulation, opting instead for oversight through voluntary agreements with tech companies. The approach is seen as shifting amid a blizzard of warnings from AI developers and executives.
On Thursday, Mustafa Suleyman, a Brit who co-founded AI company DeepMind before it was acquired by Google and now heads Microsoft AI, warned that “we are essentially seeding a new silicon species” that would compete with humans. There are also growing concerns that advanced AI models can now game independent evaluation.
London boasts a big AI presence, including the headquarters of Google DeepMind, as well as expertise nearby at leading universities such as Oxford and Cambridge. OpenAI has committed to investing in a London research hub, further adding to the city’s place as Europe’s center of gravity in the industry. Britain ranked third in 2025 behind the U.S. and China in private investment, followed by France, according to a Stanford University global industry report. The U.K. is rich in AI researchers and inventors, but in terms of churning out widely adopted AI models, China and the U.S. leave the rest of the world in the dust.
King Charles, who doesn’t use a cellphone but has repeatedly used his international standing to convene policymakers to discuss thorny topics including climate change, was tasked with trying to knock heads together. The meeting wasn’t expected to result in any binding commitments. Rather, it aimed to set a direction of travel.
The depth of feeling was on display in Scotland, where campaigners drew a sprawling message on a sandy beach just below the local airport where executives landed. It read: “AI? EH NAW” — which, translated from Scottish slang, stands for “AI? Er, no.”