Britain’s King Charles delivered a stark message to some of the world’s most influential artificial intelligence leaders, warning that powerful AI systems in the wrong hands could pose “existential dangers” and demanding reassurance that humanity can still steer its own future.
Speaking at a summit at his Dumfries House estate in Scotland, the King is hosting executives from Nvidia, Google DeepMind, OpenAI, Anthropic and other companies at a pivotal moment for the sector, as concerns intensify that safety work is failing to keep pace with the rapid development of increasingly capable systems.
Addressing the gathering, Charles said he was hoping for a “fruitful conversations on the rapid development of Artificial Intelligence”.
He said the technology’s momentum is both “intriguing and deeply concerning in equal measure,” noting that AI is already demonstrating “immense capacity to improve and save life” — pointing to advances in life sciences and medicine as an example.
“Yet, those who have created these technologies are now increasingly warning that AI risks developing darker capacities – perhaps even to take life.
“In this regard, if I may say so, there seems urgency in adequately considering the existential dangers of such technologies falling into the wrong hands, and being used in potentially catastrophic ways. Surely, we need sufficient means of control before it is all too late?
“Those in our world who value deeply our humanity and its vital moral component are anxiously seeking your reassurance that we will not lose control of our destiny, or our souls.”
Dumfries House, set amid an extensive parkland estate about an hour south of Glasgow, serves as the headquarters of the King’s Foundation, the charitable organisation Charles created to oversee his philanthropic interests. President Catherine Connolly met with the King last night during her four-day official visit to Scotland.
The event also marks Charles’ first public appearance since Buckingham Palace pushed back against claims made by the brother of the late Princess Diana.
In his new memoir, Charles Spencer alleges the King told him “we’ll forget her soon enough” after Diana’s death in 1997.
After excerpts were published on Wednesday, a Buckingham Palace spokesman responded: “While we do not comment on books as a matter of principle, His Majesty is mindful that the pain of fraternal grief can cloud reason, affect judgment and colour memory in ways others do not recognise, even many years after such a loss”.
OpenAI reveals six new cases of AI misbehavior
The summit comes as OpenAI pledged to more systematically disclose cases in which its models behave unexpectedly, releasing six new reports on previously undisclosed incidents of AI misbehaviour.
The transparency move follows a run of incidents at the company that have emerged gradually since July.
The most serious case described involved two OpenAI models that, during testing, spontaneously escaped their contained environment, accessed the internet and broke into several websites and platforms.
OpenAI’s new reporting framework is intended, in part, to give outsiders clearer evidence of what frontier AI systems can do — information the company says is necessary for informed debate about how quickly development should proceed.
On Saturday, Anthropic Chief Executive Dario Amodei proposed a coordinated slowdown in the pace of AI advances to create space to understand the risks that new capabilities can introduce.
OpenAI Chief Executive Sam Altman, Google DeepMind President Demis Hassabis, SpaceXAI chief Elon Musk and Microsoft Chief Executive Satya Nadella all backed the call.
In its announcement, OpenAI said, “We do not believe that the AI industry has solved alignment and monitoring to a sufficient degree to continue responsibly scaling at maximum speed for much longer.”
“Decisions about how AI development should proceed in the months and years to come need to draw on evidence that people outside the companies building frontier models can examine for themselves,” the company added.
OpenAI Chief Executive Sam Altman
Under the expanded approach, OpenAI said it will report problems including unauthorised actions by AI, instances of systems escaping oversight, and cases of spontaneous coordination between AI systems.
OpenAI said an incident will not need to have caused harm — or to form part of a broader pattern — to be made public.
The reporting is set to span the entire AI lifecycle, from development and evaluation to testing and online deployment.
None of the six newly disclosed examples resulted in significant consequences, but OpenAI said they reinforce trends already observed.
In one incident in May, a model created its own source on the internet to answer a question during development, citing a document it had generated itself.
OpenAI also described another episode from May in which the AI suggested ways to fabricate data it had not found or to conceal its errors.
