A Call for Control of Frontier AI Models (joint statement, September 2026)

"A Call for Control of Frontier AI Models" is a joint political statement by heads of state and government, launched on 21 September 2026 by Prime Minister Jonas Gahr Støre of Norway and President Alexander Stubb of Finland. The document is headed "September 2026" and runs to three short paragraphs followed by three numbered calls and a signatory block. It is hosted by the Norwegian government as a PDF, published as a news item by the Office of the President of Finland on 21 September (Source: presidentti.fi), and filed as a government document by the Netherlands on 22 September (Source: government.nl). It creates no obligations and names no instrument. The topic page tracking its endorsements and reception is A Call for Control of Frontier AI Models (2026).

Text

The statement opens by describing AI as "an extremely powerful technology" with the potential to improve lives, advance science and strengthen economies, and states that "the rapid development of frontier AI models poses serious risks to safety and security if not appropriately managed."

Its second paragraph gives two grounds for acting. The first is recent conduct by AI systems: "Recently we have seen capable AI systems circumventing testing safeguards, exploiting vulnerabilities and gaining unauthorized access to real-world systems." The statement does not name the systems, developers or incidents. The second is expert warning: leading scientists and executives, it says, warn that the pace of development could outpace the ability to manage emerging risks. Neither group is named.

The third paragraph sets out principles. It states that industry, governments and society "must act now," that oversight should be strengthened "without widening the gap between countries in access to the benefits of AI," that "AI must remain under human direction, oversight and control," and that AI "must be developed and used in line with international law." It then "welcome[s] the recently launched initiatives that aim to manage the risks related to frontier technologies" without identifying them, and makes three calls:

1. Companies to develop transparent safety protocols, including mandatory pre- deployment testing and independent evaluation, with qualified evaluators granted sufficient access to assess risks.
2. Governments and regional organisations to further develop and coordinate common standards, strengthen transparency — including shared reporting of serious safety incidents — and ensure that countries across all regions have access to scientific capacity, expertise and trusted evaluation.
3. UN member states to build on existing international mechanisms and explore creating an international institution, able to set standards, enable verification, and convene states when capability thresholds are crossed.

Structure of the calls

Each call is addressed to a different set of actors, and none is addressed to the signatories as governments acting individually.

• Companies. The first call asks developers to adopt "transparent safety protocols." Mandatory pre-deployment testing and independent evaluation are listed as components of those protocols; the text does not say who would make the testing mandatory or on what legal basis. The access condition — "qualified evaluators granted sufficient access to assess risks" — does not define qualification or sufficiency. See AI Pre-Release Vetting, Third-Party AI Assessment.
• Governments and regional organisations. The second call covers common standards, transparency including "shared reporting of serious safety incidents," and access across all regions to scientific capacity, expertise and "trusted evaluation." It does not specify a reporting recipient, a severity definition or a deadline. See International AI Standards, AI Incident Reporting.
• UN member states. The third call asks states to "build on existing international mechanisms," which it does not name, and to "explore creating" an international institution with three functions: setting standards, enabling verification, and convening states "when capability thresholds are crossed." The thresholds are not defined, and the text does not say whether convening would lead to any collective action. See Frontier AI Governance.

The statement does not call for a pause or slowdown of development, a point noted in coverage by The Straits Times (Source: straitstimes.com). The Norwegian government's press release accompanying the launch attributes pacing language to Støre personally: "the development of frontier AI models must take place at a pace that enables us to retain control" (Source: regjeringen.no). That sentence is not in the statement.

Signatories

The signatory block lists leaders by name and office. Most are heads of state or government; Bahrain, Türkiye and the United Arab Emirates are represented by foreign ministers (the UAE's also Deputy Prime Minister), and Liechtenstein by its Hereditary Prince. Finland is represented twice, by President Stubb and Prime Minister Petteri Orpo. The European Commission is represented by its President, Ursula von der Leyen.

The version captured in Raw Sources/ lists 28 signatories, with footnoted endorsement dates running to 23 September 2026. The Norwegian-hosted PDF retrieved again on 25 September 2026 lists 30, adding Andrej Plenković, Prime Minister of Croatia, and Luis Montenegro, Prime Minister of Portugal (Source: regjeringen.no). Its dated endorsement notes read:

Both versions state that "the statement remains open for endorsement from leaders of other countries." The 30 signatories of 24 September, in the document's order, are: Jonas Gahr Støre (Norway); Alexander Stubb and Petteri Orpo (Finland); Anthony Albanese (Australia); Christian Stocker (Austria); Dr. Abdullatif bin Rashid Al Zayani (Bahrain); Mark Carney (Canada); Andrej Plenković (Croatia); Mette Frederiksen (Denmark); Kristen Michal (Estonia); Ursula von der Leyen (European Commission); Emmanuel Macron (France); Friedrich Merz (Germany); Kristrún Frostadóttir (Iceland); Micheál Martin (Ireland); Kassym-Jomart Tokayev (Kazakhstan); William Ruto (Kenya); Edgars Rinkēvičs (Latvia); Prince Alois (Liechtenstein); Luc Frieden (Luxembourg); Maia Sandu (Moldova); Rob Jetten (Netherlands); Luis Montenegro (Portugal); Nicușor Dan (Romania); Lawrence Wong (Singapore); Julius Maada Bio (Sierra Leone); Pedro Sanchez (Spain); Cyril Ramaphosa (South Africa); Hakan Fidan (Türkiye); and Sheikh Abdullah bin Zayed Al Nahyan (United Arab Emirates). The Finnish presidency's web version lists the same 30 names, with Stubb first (Source: presidentti.fi).

No leader of the United States, the United Kingdom or China appears on either version; The Straits Times reported the absence of the United States and China on 22 September. Of the G7, Canada, France and Germany signed; the United States, the United Kingdom, Japan and Italy did not.

Discrepancies noted

• The Norwegian Prime Minister's Office press release of 21 September states that "22 leaders from around the world support the appeal" (Source: regjeringen.no). The Straits Times list of 22 September names Singapore, the EU and 19 other countries, which matches 22 individual leaders because Finland has two. Austria and Romania, recorded in the statement as endorsing on 21 September, are absent from both.
• The capture held in Raw Sources/ is an earlier state of the same PDF: its 23 September note names France and Sierra Leone, whereas the version retrieved on 25 September names Croatia, France and Sierra Leone for that date. The raw file's own frontmatter describes endorsement by "some thirty states," but the capture lists 28 individuals from 26 states and the European Commission.
• A New Developments Log digest of 25 September 2026 reports "28 signatories as of its September 24 update" (Source: New Developments Log/2026-09-25-0505-ai-developments.md). The version carrying the 24 September endorsement note lists 30.

Provenance

The raw file was captured on 25 September 2026 by gap-scan from the Norwegian government's canonical host (regjeringen.no) with the PDF parser, and corroborated against the Government of the Netherlands document page, the Finnish presidency page, and an ANI report of the UN Secretary-General's response. The PDF's signatory block is set in columns and extracts as run-together names and titles; the capture does not reorder them. The PDF was retrieved again on 25 September 2026 for this page, returning the later 30-signatory version (Source: regjeringen.no). The operative text is identical in both versions and on the Finnish presidency page, apart from the line break inside "pre- deployment" in the PDF.

Relationships

• instance-of: Frontier AI Governance — an intergovernmental statement on control of frontier models.
• supports: AI Pre-Release Vetting, Third-Party AI Assessment — calls for mandatory pre-deployment testing and independent evaluation with evaluator access.
• supports: AI Incident Reporting — calls for shared reporting of serious safety incidents among governments.
• supports: International AI Standards — calls for coordinated common standards and an international standard-setting institution.
• related: A Call for Control of Frontier AI Models (2026) — topic page: endorsements, context and reception.
• related: Jonas Gahr Støre, Alexander Stubb — initiators.
• related: António Guterres — welcomed the statement on 21 September 2026.
• related: Coordinated slowdown proposals compared — the statement sits among the September 2026 governmental responses to calls for pacing.
• related: AI Agent Containment Incidents, 2026 — the incidents of AI systems gaining unauthorized access to real-world systems that the statement cites without naming.
• related: AI Control, Human Oversight.
• related: Bletchley Declaration (AI Safety Summit, 2023), Paris AI Action Summit Declaration (2025) — earlier multilateral AI declarations.