# Council of Europe AI Convention

The Council of Europe AI Convention, formally the Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law, was opened for signature in 2024, establishing the first binding international treaty on AI.

The Council of Europe AI Convention, formally the Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law, is an international treaty opened for signature on 5 September 2024 in Vilnius, Lithuania. It represents the first legally binding instrument specifically addressing artificial intelligence, aiming to ensure that AI systems are developed and used in ways consistent with human rights, democratic processes, and the rule of law. The convention applies to the activities of public authorities and private actors acting on their behalf, with provisions for risk assessment and mitigation throughout the lifecycle of AI systems.

The treaty emerged from a multi-year process within the Council of Europe, which began in 2019 with the establishment of the Ad Hoc Committee on Artificial Intelligence (CAHAI). After CAHAI’s feasibility study, the Committee on Artificial Intelligence (CAI) was tasked with drafting the convention in 2022. Negotiations concluded in March 2024, and the text was adopted by the Committee of Ministers on 17 May 2024. The convention is open to non-European states as well, reflecting its global ambition. As of early 2025, several countries, including the United Kingdom, the United States, and Japan, have signed but not yet ratified the treaty.

## Background and Context

The rapid advancement of [artificial intelligence](https://www.wikiprompt.org/wiki/artificial-intelligence) technologies, including [machine-learning](https://www.wikiprompt.org/wiki/machine-learning) and [deep-learning](https://www.wikiprompt.org/wiki/deep-learning) models, prompted international efforts to establish common norms. The Council of Europe, with its long-standing focus on human rights, positioned itself as a key forum for such regulation. The convention builds on existing instruments like the European Convention on Human Rights and the Convention for the Protection of Individuals with regard to Automatic Processing of Personal Data. Unlike the [European Union’s AI Act](https://www.wikiprompt.org/wiki/eu-ai-act-council-approval), which is a regional regulation, the Council of Europe convention is designed to be a global framework, open to countries beyond Europe.

The drafting process involved extensive consultation with civil society, academia, and industry. Key figures in the negotiations included the chair of the CAI, and representatives from member states and observer states. The convention’s scope covers the entire lifecycle of AI systems, from design and development to deployment and decommissioning, with a focus on activities that may interfere with human rights.

## Key Provisions

The convention establishes general principles for AI systems, including transparency, accountability, and non-discrimination. It requires parties to adopt measures to ensure that AI systems are used in a manner consistent with human dignity and individual autonomy. Specific provisions address the protection of personal data, the right to privacy, and the right to effective remedy for those adversely affected by AI decisions. The treaty also emphasizes the importance of public awareness and education about AI.

One of the convention’s core elements is the requirement for risk assessment and mitigation. Parties must identify, assess, and address risks posed by AI systems, particularly those that could lead to discrimination or other human rights violations. The convention distinguishes between high-risk and low-risk applications, with stricter obligations for the former. It also mandates that AI systems be subject to human oversight, ensuring that humans can intervene in decision-making processes when necessary.

## Signatories and Ratification Process

As of early 2025, the convention has been signed by over 20 countries, including the United Kingdom, the United States, Japan, and several European nations. The signing ceremony took place in Vilnius, with representatives from these countries in attendance. Ratification requires domestic approval in each signatory state; the convention will enter into force three months after the fifth ratification, including at least three Council of Europe member states. As of the knowledge cutoff, no country has yet ratified, but the process is underway in several jurisdictions.

The convention’s open nature allows non-member states to accede, which has been seen as a way to promote global standards. For instance, the United States, which is not a member of the Council of Europe, has signed the treaty, signaling its willingness to align with international norms on AI governance. However, ratification is not guaranteed, and the treaty’s effectiveness will depend on the commitment of signatories to implement its provisions.

## Comparison with the EU AI Act

The Council of Europe AI Convention and the [EU AI Act](https://www.wikiprompt.org/wiki/eu-ai-act-council-approval) are complementary but distinct instruments. The EU AI Act is a comprehensive regulation that applies to all AI systems placed on the EU market, with a risk-based classification and binding requirements for providers and users. In contrast, the Council of Europe convention is a framework treaty that sets general principles and obligations, leaving implementation details to national legislation. The convention is less prescriptive, allowing parties flexibility in how they achieve its goals.

Both instruments emphasize human rights and safety, but the convention has a broader geographic reach, potentially influencing AI governance beyond Europe. The EU AI Act, which was approved by the European Council in May 2024, is expected to serve as a model for many countries, but the Council of Europe treaty offers an alternative or complementary path for non-EU states. The relationship between the two is not hierarchical; rather, they operate in parallel, with the EU AI Act applying to EU member states and the convention providing a global baseline.

## Implementation and Enforcement

The convention requires each party to establish or designate a supervisory authority to oversee compliance. These authorities must have the power to investigate, impose sanctions, and provide remedies for violations. The treaty also encourages international cooperation, including mutual assistance and exchange of information between authorities. However, enforcement mechanisms are relatively weak compared to the EU AI Act, which includes fines of up to 7% of global turnover for certain violations.

Implementation will require significant legal and institutional changes in many countries. For example, parties must ensure that their domestic laws align with the convention’s principles, which may involve amending existing data protection and anti-discrimination laws. The convention also calls for the development of standards and guidelines, possibly in collaboration with other international bodies like the OECD and UNESCO.

## Reactions and Criticisms

The convention has received mixed reactions. Supporters, including many human rights organizations, praise it as a historic step toward binding international norms for AI. They argue that it provides a necessary counterbalance to the rapid deployment of AI systems by companies like [openai](https://www.wikiprompt.org/wiki/openai), [anthropic](https://www.wikiprompt.org/wiki/anthropic), and [google-deepmind](https://www.wikiprompt.org/wiki/google-deepmind), which are developing increasingly powerful [large language models](https://www.wikiprompt.org/wiki/large-language-model). Critics, however, contend that the convention is too vague and lacks enforcement teeth. Some civil society groups have called for stronger provisions on bans of certain AI applications, such as social scoring systems, and for clearer liability rules.

Industry representatives have expressed concerns about the potential for regulatory fragmentation, as different countries may interpret the convention differently. They also worry about the compliance burden, particularly for small and medium-sized enterprises. Nevertheless, many tech companies have publicly supported the convention’s goals, seeing it as a way to build public trust in AI.

## Future Outlook

As of early 2025, the convention is in the ratification phase. Its entry into force is expected within the next few years, depending on the speed of ratifications. The treaty’s impact will likely be significant, as it sets a precedent for international AI governance. It may also influence the development of other regional frameworks, such as those in Africa and Latin America. The convention’s success will depend on how effectively parties implement its principles and whether it can adapt to the evolving AI landscape, including advances in [generative-ai](https://www.wikiprompt.org/wiki/generative-ai) and autonomous systems.

The Council of Europe plans to hold regular meetings of the parties to review implementation and share best practices. There is also a possibility of additional protocols to address specific issues, such as AI in the military domain or in healthcare. The convention is a living instrument, designed to evolve with technological developments.

## Conclusion

The Council of Europe AI Convention marks a milestone in the global governance of artificial intelligence. By establishing binding obligations on states to protect human rights in the AI context, it sets a new standard for international cooperation. While its effectiveness remains to be seen, the convention represents a collective effort to ensure that AI serves humanity rather than undermines it. As countries move toward ratification, the treaty’s principles are likely to shape national laws and corporate practices worldwide, contributing to a more accountable and human-centric AI ecosystem.

---
Source: https://www.wikiprompt.org/wiki/council-of-europe-ai-convention
License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/)
Last updated: 2026-11-03T12:40:22.045502+00:00