Ireland’s government says the Dublin Declaration was adopted on October 14, 2026, with seven priorities for AI policy. It is a political declaration; Norway’s separate call asks companies and governments to take specific safety measures, while the Council of Europe’s Framework Convention is a treaty whose entry-into-force threshold remained unmet in the Treaty Office’s October 23 status table.
Three initiatives, three different commitments
The instruments differ in both their scope and their legal form. Norway’s document is a call for action, Ireland describes Dublin as a shared commitment, and the Council of Europe convention is a treaty intended to be legally binding for its parties.
| Instrument | What it addresses | Legal position |
| Norway’s A Call for Control of Frontier AI Models | Company safety protocols and testing; government coordination and incident reporting; exploration of an international institution | A political call, not a treaty |
| Dublin Declaration on Trustworthy, Safe and Sovereign Artificial Intelligence | Seven priorities spanning safety, AI adoption, sovereignty, infrastructure, sustainability and cooperation | A political declaration adopted by participating governments |
| Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law | Human rights, democracy and rule-of-law safeguards across AI activities, with risk assessments and monitoring | A treaty; the entry-into-force threshold had not been met by October 23, 2026 |
What Norway’s call asks companies and governments to do
Published on September 21, 2026, Norway’s call for control of frontier AI models asks AI companies to adopt transparent safety protocols, conduct mandatory tests before deployment and arrange independent evaluations.
It also asks governments to coordinate on standards, transparency and the reporting of serious safety incidents. The statement invites UN member states to explore an international institution that could set standards, support verification and convene states when AI capabilities reach specified thresholds. Those are requests in a political call, not obligations created by a treaty.
Norway’s notice lists 30 national leaders and the President of the European Commission. It records endorsement updates through October 22 and says the call remains open to further endorsements. The proposal’s emphasis on outside scrutiny connects to a separate debate over who should assess AI companies’ safety practices, discussed in NeoTeo’s earlier coverage of external oversight.
The Dublin Declaration’s seven priorities
Ireland’s government says the declaration was adopted on October 14, 2026. Its priorities extend beyond model safety to the conditions for developing and using AI across Europe:
- AI safety and resilience.
- Trustworthy AI adoption in public services and industry.
- European AI sovereignty.
- Infrastructure, skills and computing capacity.
- Sustainability.
- Public-sector innovation.
- European and international cooperation.
Ireland presents the declaration as a shared political commitment. Its status is distinct from that of a ratified treaty.
What the Council of Europe treaty covers
The Framework Convention covers AI used by public authorities, private actors working on their behalf, and private actors. For rules governing private-sector activity, parties may apply relevant treaty provisions directly or use other measures that meet the convention’s requirements while respecting human rights, democracy and the rule of law.
Its safeguards include human dignity and autonomy, equality, privacy and data protection, transparency, accountability and reliability. It calls for iterative risk and impact assessments, mitigation measures, procedural safeguards and ways to complain. A Conference of the Parties provides a monitoring mechanism.
The convention also sets out exceptions. Parties are not required to apply it to activities protecting national-security interests, although those activities must respect international law and democratic institutions and processes. National defence is excluded. Research and development is also excluded, except where testing could interfere with human rights, democracy or the rule of law.
Treaty status and the United States
The Council of Europe Treaty Office’s status table dated October 23, 2026 recorded one ratification or accession toward the five required for the convention to enter into force. At least three of the five must come from Council of Europe member states. The European Union ratified the convention on May 15, 2026.
The United States signed the convention on September 5, 2024, but had not ratified it by October 23, 2026. A signature and a ratification are separate steps; the treaty had not reached its entry-into-force threshold in that dated status table.
