The Council of Europe’s framework convention on artificial intelligence was opened for signature in September 2024 and is widely described in international legal literature as the first binding international AI treaty. The instrument aims to ensure AI development and use do not conflict with human rights, democracy and the rule of law. When the treaty opened, it received prominent political backing: the European Union, the United States and the United Kingdom were among the first signatories, and Canada and Japan also signed.
Why had the treaty not entered into force by September 2026?
According to analysis by international lawyer Mujib Jimoh, by September 2026 the convention had not yet entered into force. Entry into force requires five ratifications, including ratification by at least three Council of Europe member states; the Council of Europe’s ratification table at that time showed the European Union was the only actor that had actually deposited the required instrument.
It is important to distinguish signing from ratifying: signing indicates political support for the treaty’s objectives but does not make the signatory a legally bound party. Ratification creates legal obligations and typically requires domestic approval processes such as parliamentary decisions, interministerial coordination and legal conformity reviews.
"Expressive hard law": hard form, initially soft operation
Jimoh characterizes the convention as "expressive hard law." Formally it is a treaty—not a recommendation or ethical code—but in practice its mechanisms have so far been more symbolic: institutional enforcement and interpretation processes have not yet started. Through signing, states gain reputational benefits by publicly supporting human-rights-based AI regulation, while signing alone does not force immediate, comprehensive domestic legal changes.
At the same time, the Vienna Convention’s Article 18 obliges signatory states, even before ratification, to refrain from acts that would defeat the object and purpose of the treaty. Thus signing is less than full legal commitment but more than a mere public gesture.
Two main reasons for slow ratification
There are several complementary explanations for the delayed ratification:
- Practical domestic procedures: in many countries signing is administratively simpler than completing the ratification process, which can require legislative approval and detailed legal alignment.
- Political and industrial caution: if the treaty enters into force, the Conference of the Parties would convene to interpret the convention, monitor implementation and gradually shape common expectations about what human-rights-based AI regulation means in practice. Because AI has become an industrial policy and national security issue, many governments are wary of international frameworks that could constrain domestic policy space—particularly if competitors do not take on comparable obligations.
The European Union’s position and domestic legal groundwork
The European Union was able to move sooner because it already adopted its detailed AI regulation, the EU AI Act. The EU approach envisages implementing the Council of Europe convention’s requirements primarily through this regulation and other EU law. For businesses operating in or targeting the EU market, this translates into concrete compliance duties: documentation, risk management, transparency obligations, data quality and governance requirements.
Regulatory fragmentation and corporate pain points
The global regulatory environment remains fragmented. Some regions have detailed statutes, others rely on softer guidance, national strategies or sectoral rules. Jimoh notes that the problem for companies is not the absence of rules but the coexistence of many different regulatory logics running in parallel.
At the same time, a converging set of principles is emerging internationally—fairness, accountability, transparency and privacy protection—which are essentially human-rights requirements translated into the AI context.
Conclusion: not a failure, but not yet underway
By September 2026 the world’s first binding international AI convention had not failed, but it had not yet progressed into operational effect either. It functions largely as a mirror for states: many governments endorse human-rights-based AI regulation in principle, but their enthusiasm cools when it comes to shared international monitoring and interpretation. Mujib Jimoh’s essay is part of a broader research project examining the international law and customary law questions raised by artificial intelligence.
Author and background
The underlying analysis was prepared by Mujib Jimoh, an international lawyer and analyst at Meridiam Infrastructure North America Corp. He holds an LL.M. from Duke University School of Law, a primary law degree from the University of Ibadan, and is qualified to practise law as a barrister and solicitor in Nigeria.
