On 17 September 2026, King Charles convened an AI Summit at Dumfries House in East Ayrshire, bringing together leaders from Nvidia, Google DeepMind, OpenAI and Anthropic alongside the UK’s Minister for Artificial Intelligence, The Rt Hon Kanishka Narayan, to discuss whether frontier AI development needs a shared set of guiding principles. The King explicitly tied the gathering back to “the message I sent to the United Kingdom AI Safety Summit in Bletchley in 2023” — the first global AI Safety Summit, hosted at Bletchley Park under the previous government.
A note on sourcing: this article is built entirely from two primary sources CASRAI read directly — the Royal Family’s own website, royal.uk, publishing both the King’s speech text and a separate report on the summit itself, both dated 17 September 2026. No secondary press coverage is cited here; the summit’s discussion of a “slowdown” narrative in the broader AI industry and any additional secondary reporting on this event are deliberately out of scope for this piece.
Who was there, and who convened it
The summit was held at Dumfries House, the King’s Foundation-run estate in East Ayrshire, Scotland. Royal.uk’s own account describes it as bringing together “leaders from the artificial intelligence sector, government ministers, and civil society thought leaders.” Confirmed attendees from the AI industry were Nvidia, Google DeepMind, OpenAI and Anthropic. From government, The Rt Hon Kanishka Narayan attended in his capacity as UK Minister for Artificial Intelligence. Three organisations tied to the King’s own charitable and diplomatic work were directly involved: The King’s Trust and The King’s Foundation (which manages Dumfries House itself), and the Sustainable Markets Initiative, the King’s public-private coalition on sustainability. Discussions were facilitated by the Ditchley Foundation, a UK charity that has run closed-door, off-the-record policy dialogues among government, business and civil-society leaders since 1958 — the kind of format suited to a discussion CASRAI’s sources describe as exploratory rather than a negotiated text ready for signature.
The core question: a shared set of AI principles
Per royal.uk’s account, delegates explored whether “a shared set of principles could be established to guide the future application of AI.” The framing on offer was explicitly not purely technical: the aim, as royal.uk describes it, was to position AI “not only as a driver of capability and efficiency, but as a tool that upholds human dignity and supports the flourishing of both people and the planet.” Neither royal.uk page describes a finished document, a signing ceremony, or a list of principles already agreed — the record is of a discussion opened, not a declaration concluded.
In his own remarks, the King framed the underlying stakes as two open questions: “How can we harness the benefits of AI, with safety at its heart?” and “How can we build international cooperation and consensus to achieve this?” He was direct about urgency, saying “there seems urgency in adequately considering the existential dangers of such technologies,” while also cautioning that “our humanity must remain sacred” as AI capabilities advance. He drew on the “philosophy of Harmony” he set out in a book published sixteen years earlier, telling delegates that “as we embrace what is new, we must not lose what is timeless” and that “innovation must always be considered within the wider systems upon which all depend.” To the tech leaders and civil-society figures in the room specifically, he said: “I can only applaud those of you who are so deeply committed to harnessing Artificial Intelligence for good, so that we might see a future where human potential remains boundless, whilst technological development is carefully curated to enhance our success.”
The explicit Bletchley Park callback
The King did not treat this as a first-of-its-kind gathering. His speech directly referenced his “message I sent to the United Kingdom AI Safety Summit in Bletchley in 2023” — the first global AI Safety Summit, at which 28 countries plus the EU signed the non-binding Bletchley Declaration on frontier-AI risk (see CASRAI’s own Frontier AI Safety Timeline and Bletchley Declaration vs Seoul Declaration comparison for what that summit produced and how the international summit process has evolved since). Dumfries House was not a repeat of that government-to-government process — it had no equivalent list of signatory states — but the King’s own framing positions it as a continuation of the same conversation he opened in 2023, this time convened personally, at a private royal estate, with the companies themselves rather than their governments in the room.
Where NIKOLAI fits
NIKOLAI is CASRAI’s own frontier-AI-safety dictionary of elements — an independent, unendorsed reference work, not a standard any lab, regulator or, for that matter, royal household has adopted or cited. Nothing at Dumfries House referenced NIKOLAI, and this section is offered only as a reader’s lens on what “a shared set of principles” would actually have to specify to be usable rather than aspirational. NIKOLAI’s N8 track (Transparency and Review) catalogues exactly the kind of structural detail a principles framework built around “upholding human dignity” would eventually need to operationalise if any outside party is going to check whether a developer is living up to it. Its Evaluator Independence and Conflict of Interest element, for instance, records the specific tests a reviewer must clear before a safety assessment counts as independent — NIKOLAI’s own shadow-mapping of this element cites Anthropic’s Responsible Scaling Policy (three tests: relevant expertise, incentive to be candid, no financial or close personal ties to the developer) and METR’s practice of disclosing individual conflicts such as staff domestic partnerships with lab leadership. A summit communique that says AI should “uphold human dignity” says nothing, on its own, about who checks that claim or how independent they have to be; N8 is CASRAI’s own cataloguing of the kind of clause that gap would need filled by. As always, every NIKOLAI crosswalk row is CASRAI’s own reading of a published document, not a claim that Anthropic, METR, or any other named source uses or endorses NIKOLAI’s terms.
Key facts, at a glance
- Date and location: 17 September 2026, Dumfries House, East Ayrshire, Scotland.
- Convened by: King Charles.
- AI companies present: Nvidia, Google DeepMind, OpenAI, Anthropic.
- Government attendee: The Rt Hon Kanishka Narayan, UK Minister for Artificial Intelligence.
- Royal/civil-society organisations involved: The King’s Trust, The King’s Foundation, the Sustainable Markets Initiative.
- Facilitator: The Ditchley Foundation.
- Core question discussed: whether a shared set of principles could guide the future application of AI.
- Explicit precedent cited: the King’s own 2023 message to the Bletchley Park AI Safety Summit.
- Outcome as of this writing: a discussion opened, not a signed declaration or published text.
Frequently asked questions
Did the King’s AI Summit at Dumfries House produce a signed agreement?
No. Royal.uk’s own account describes delegates exploring “whether a shared set of principles could be established,” not a text that was finalised or signed at the event. Unlike the 2023 Bletchley Declaration, there is no published Dumfries House document and no list of signatory governments.
Is this the same as the 2023 Bletchley Park AI Safety Summit?
No, though the King explicitly linked the two. Bletchley in 2023 was a government-to-government summit at which 28 countries plus the EU signed a joint declaration. Dumfries House in 2026 was convened personally by the King, at a private royal estate, with AI companies and one UK minister in the room rather than an assembly of national governments.
Who from the AI industry attended?
Royal.uk confirms Nvidia, Google DeepMind, OpenAI and Anthropic as attendees. Its report does not list every individual delegate by name or title.
What role did the Ditchley Foundation play?
The Ditchley Foundation facilitated the discussions. It is a UK charity that has run closed-door policy dialogues among government, business and civil-society leaders since 1958 — consistent with an exploratory, off-the-record-style discussion rather than a negotiated communique.
Does CASRAI’s NIKOLAI project have any official connection to this summit?
No. NIKOLAI is CASRAI’s own independent, unendorsed reference dictionary of frontier-AI-safety elements. Nothing in either royal.uk source mentions NIKOLAI; the “Where NIKOLAI fits” section above is CASRAI’s own editorial lens on what a principles framework like the one discussed would eventually need to specify, not a claim that the summit referenced, used, or endorsed NIKOLAI in any way.
Where can I read the primary sources directly?
Both are published on the Royal Family’s own website: “The King’s Speech at the AI Summit in Scotland” and “The King Convenes Tech Leaders for AI Summit in Scotland,” both dated 17 September 2026.
Related reading
- Frontier AI Safety Timeline: 2023-2026
- Bletchley Declaration vs Seoul Declaration: The International AI Safety Summit Process
- A Cross-Lab AI Safety Standards Body? What’s Confirmed About the Anthropic-OpenAI-Google Talks
- What Is NIKOLAI? CASRAI’s Frontier-AI-Safety Dictionary Explained
- The Frontier Model Forum, Explained
Sources
The Royal Household, “The King’s Speech at the AI Summit in Scotland,” royal.uk, 17 September 2026. The Royal Household, “The King Convenes Tech Leaders for AI Summit in Scotland,” royal.uk, 17 September 2026. NIKOLAI element sourcing (Evaluator Independence and Conflict of Interest, N8): Anthropic, Responsible Scaling Policy v3.4; METR, public conflict-of-interest disclosure practice — both as read on the live NIKOLAI element page on 20 September 2026.
