Thinking Thinking
Bundle of French tricolour flags on gold-tipped poles mounted on the stone facade of the Élysée Palace in Paris, between two tall windows with wrought iron balustrades
Foto: Netfalls Remy Musser/Shutterstock.com

News

Paris Call on AI safety: Luxembourg signs up

Luxembourg is among 17 states and ten AI firms, including Mistral AI, OpenAI, Anthropic and Microsoft, behind the non-binding Paris Call for AI Safety and Security of 09.11.2026.

By Nova 4 min read

Key takeaways

  • Luxembourg is one of 17 states on the Élysée's signatory list, read on 12.11.2026.
  • Ten AI companies signed, including Anthropic, Microsoft, Mistral AI and OpenAI; Google, Meta and xAI are not listed.
  • Neither the European Commission nor the EU AI Office is a signatory.
  • The declaration is not legally binding; in the EU the AI Act remains the binding framework.
On this page
  1. Who signed the Paris Call for AI Safety and Security?
  2. What does the Paris Call commit signatories to?
  3. Did the European Commission or the AI Office sign?
  4. What does the Paris Call mean for companies in Luxembourg?
  5. How much will the Paris Call change?

Luxembourg is one of 17 states on the current signatory list of the Paris Call for AI Safety and Security, adopted at the Élysée Palace on 09.11.2026.

Ten AI companies also signed, among them Mistral AI, OpenAI, Anthropic and Microsoft. The declaration is not legally binding.

Who signed the Paris Call for AI Safety and Security?

The English text on elysee.fr, read on 12.11.2026, lists Austria, Bahrain, Canada, Croatia, Cyprus, Estonia, Finland, France, Germany, Ireland, Kenya, Latvia, Liechtenstein, Luxembourg, the Netherlands, Norway and Portugal. The companies are Anthropic, Black Forest Labs, Cohere, Domyn, Hugging Face, Microsoft, Mistral AI, OpenAI, Sakana AI and Synthesia, with The Future Society also named in that group. Civil society signatories include Yoshua Bengio's LawZero, SaferAI, Ifri and the French centre CeSIA.

The list keeps growing. The Canadian government's statement of 09.11.2026 named only six states and did not include Luxembourg. The Élysée does not say when Luxembourg joined. Versions also differ: the Dutch government's copy names Japan, Sierra Leone and the United Kingdom, which are not on the Élysée list. Google, Meta and xAI appear on none of the lists.

What does the Paris Call commit signatories to?

The text has three pillars. First, policymakers, researchers and developers aim to build a common framework for evaluating and managing the risks of frontier models, focused on misuse in the cyber and CBRN domains and on loss of control. Non-sensitive evaluation results are to be published in a standard format in a public repository. Second, developers commit to disclose their models' capabilities, limitations and known risks, and to build fail-safe mechanisms that can restrict, pause or shut down a model. Third, signatories back early warning systems for serious cyber and AI control incidents and secure channels for whistleblowers.

Many pledges are hedged. Independent evaluation is supported only "where appropriate and proportionate to the risk", and a cooperation mechanism on incidents is still to be explored. CeSIA signed but criticises the lack of common risk thresholds and of powers for public authorities to suspend a model.

Weekly, Thursday morning, Central European Time.

The Thinking briefing

One email a week: what changed in European business software, which prices moved, and which comparison we published. No sales mail.

We store your address to send the briefing. One click unsubscribes you, and we never pass it on.

Did the European Commission or the AI Office sign?

No, according to the primary sources. Neither the Commission nor the EU AI Office is on the signatory list, and the Élysée statement on the meeting names the UN and the OECD as participants, not the EU. Commission President Ursula von der Leyen did endorse the Call for Control of Frontier AI Models launched by Norway and Finland on 21.09.2026, which the Paris Call cites. Prime Minister Luc Frieden endorsed that call on 22.09.2026.

What does the Paris Call mean for companies in Luxembourg?

Nothing changes right away: the Paris Call creates no obligations, deadlines or penalties. The binding framework remains the EU AI Act. Its duties for providers of the most advanced models have applied since 02.08.2025, and since 02.08.2026 the AI Office can, according to the AI Act Service Desk, request information and model access and impose fines of up to 3 percent of global annual turnover. Anthropic, Microsoft, Mistral AI and OpenAI have also signed the General-Purpose AI Code of Practice. Under the AI Omnibus, the Annex III high-risk rules apply only from 02.12.2027.

For Luxembourg firms that use these models, the call is mainly a signal that providers promise more disclosure on capabilities and risks. That information can feed into an internal AI usage policy or a test in an AI regulatory sandbox.

As of 12.11.2026 we found no statement from the Luxembourg government on the Paris Call on gouvernement.lu. The topic stays on the agenda: according to the government, Frieden attends an informal European Council on AI in Dublin on 13.11.2026, which will also cover AI safety and governance.

How much will the Paris Call change?

The Paris Call lines up governments and major labs behind shared goals for safer frontier models. Whether it becomes more than a statement of intent depends on whether the evaluation framework, the public repository and the warning systems are actually built. In the EU, the AI Act still sets the rules.

Questions and answers

Is the Paris Call legally binding?
No. It sets out shared commitments but creates no obligations, deadlines or penalties. In the EU, the AI Act remains the binding law.
Did the European Commission sign the Paris Call?
No. The Commission and the AI Office are not on the signatory list. Ursula von der Leyen did endorse the earlier Call for Control of Frontier AI Models on 21.09.2026.

Sources

Every price and fact above was checked against these pages on the date shown.

  1. [1] Paris Call for AI safety and security (Élysée, English) accessed 2026-11-12
  2. [2] Appel de Paris pour la sécurité de l'intelligence artificielle (Élysée) accessed 2026-11-12
  3. [3] Communiqué : réunion IA et sécurité (Élysée) accessed 2026-11-12
  4. [4] Joint declaration on Paris call for AI safety and security (Prime Minister of Canada) accessed 2026-11-12
  5. [5] Paris Declaration on AI safety and security (Government of the Netherlands) accessed 2026-11-12
  6. [6] A Call for Control of Frontier AI Models (Government of Norway) accessed 2026-11-12
  7. [7] Paris Call for AI Safety: CeSIA signs, urges safeguards accessed 2026-11-12
  8. [8] Luc Frieden au 9e Sommet de la Communauté politique européenne et à la réunion informelle du Conseil européen à Dublin accessed 2026-11-12
  9. [9] Commission enforcement powers for the most advanced models from 2 August 2026 (AI Act Service Desk) accessed 2026-11-12
  10. [10] AI Omnibus enters into force (European Commission) accessed 2026-11-12
  11. [11] The General-Purpose AI Code of Practice (European Commission) accessed 2026-11-12
Topics ai-safetyparis-callluxembourgai-actfrontier-models