The Artificial Intelligence Framework Convention: What Impact on Human Rights?

The Council of Europe has achieved a world first with the opening for signatures of the Framework Convention on Artificial Intelligence and Human Rights, Democracy, and the Rule of Law. The AI Convention is open to all states and has already attracted the signatures of the United States, Israel, and also the European Union. The Lord Chancellor, Shabana Mahmood, signed on behalf of the UK.

Overview and Objectives

The Convention’s lengthy preamble recognises both the opportunities and risks that AI presents – to democracy, human rights, and the rule of law. It recalls the European Convention on Human Rights (“ECHR”), as well as the two centremost UN treaties – the International Covenant on Civil and Political Rights (“ICCPR”), and International Covenant on Economic, Social and Cultural Rights (“ICESCR”), and several specialist treaties (e.g. on children’s rights and the rights of persons with disabilities).

The ”object and purpose” of the treaty is set out in Article 1(1):

“To ensure that activities within the lifecycle of artificial intelligence systems are fully consistent with human rights, democracy and the rule of law…"

Article 1(2) goes on to require parties to “adopt or maintain appropriate legislative, administrative or other measures to give effect to the provisions set out in this Convention”. Article 1(3) refers to a “follow-up mechanism” to ensure effective implementation.

Scope of the Convention

Article 2 of the treaty adopts a broad definition of “artificial intelligence system” (which is almost identical to that in the European Union’s own AI Act):

“… a machine-based system that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations or decisions that may influence physical or virtual environments. Different artificial intelligence systems vary in their levels of autonomy and adaptiveness after deployment…”

The Convention applies principally to public authorities, and to private actors acting on behalf of public authorities (Article 3(1)(a)), with Parties given the option to specify how it will apply to private actors more generally (Article 3(1)(b)). A further, potentially significant, carve out is the complete exclusion of AI used in relation to “the protection of… national security interests…” (Article 3(2)). Moreover, matters “relating to national defence do not fall within the scope of this Convention”.

Chapter II of the Convention imposes “general obligations” to ensure that “activities within the lifecycle of artificial intelligence systems” are “consistent with” human rights obligations in international and national law (Article 4) and that AI systems are not used to undermine “democratic institutions and processes” (Article 5(1)). Particular attention is paid to “fair access to and participation in public debate” and the individual’s “ability to freely form opinions” (Article 5(2)).

During the drafting of the Convention the European Network of National Human Rights Institutions (“ENNHRI”) expressed concerns about the differential treatment of private actors, the national security/defence exclusions, and the degree of imprecision in some of the language. The UK’s Equality and Human Rights Commission (“EHRC”) is a member of the Network and has regarded artificial intelligence as a “priority” since 2022.

Chapter III of the Convention then sets out “principles”. These include very broad principles such as “human dignity and individual autonomy” (Article 7), and somewhat more specific ones such as on “privacy and personal data protection” (Article 11). The requirements, generally, are to “adopt or maintain measures” with significant scope for each jurisdiction to devise its own approach.

Articles 14 and 15 deal with “remedies” and “procedural safeguards” respectively. 

In relation to remedies, the Convention requires “accessible and effective remedies for violation of human rights”, as well as measures on the provision of information about AI systems, and “an effective possibility for persons concerned to lodge a complaint to competent authorities”. These requirements could be said to be met in the UK, for example, by the Human Rights Act 1998 – at least in relation to public authorities.

In relation to procedural safeguards, the Convention requires “effective procedural guarantees, safeguards and rights”. It also requires that Parties to the Convention shall “seek to ensure that, as appropriate for the context, persons interacting with artificial intelligence systems are notified that they are interacting with such systems rather than with a human”. There are additional provisions in relation to “public consultation” (Article 19) and “digital literacy and skills” (Article 20). The language, however, is largely that of encouragement and promotion rather than precisely drawn legal rights.

Ratification and Implementation

The Convention will enter into force when there have been five ratifications (including at least 3 Council of Europe States). From the UK’s point of view, ratification will follow the Convention being laid before the Houses of Parliament. 

The King’s Speech included an announcement for legislation on “the most powerful AI models”. In this regard, the UK is playing catch up in comparison with the European Union, which has already adopted its AI Act.

In terms of the UK’s approach to the AI Convention, the following will merit scrutiny:

• Ratification: although an early signatory, the UK will not be bound by the treaty until it is ratified. This will require the Government to lay the treaty before Parliament. Some states, such as the US, have signed but may never ratify the AI Convention, because of the challenges that any US administration faces to ratify treaties in accordance with its own Constitution;
• Application to private actors: although the AI Convention requires its application to public authorities, and private actors acting on their behalf (a formulation not dissimilar to how the Human Rights Act 1998 applies horizontally), it is open to the UK to decide how it will apply to private actors when not acting on behalf of a public authority;
• Designation of mechanism: the Convention requires the establishment or designation of one or more “effective mechanisms to oversee compliance”. It remains to be seen how the Government fulfils this requirement.

Impact on Human Rights?

In some respects the AI Convention is more an agreement to act than an action in itself. It is not, for example, directly comparable to the ECHR, which sets out specific rights – interference with which must proportionately address a legitimate aim with a legal basis. The AI Convention does not contain specific rights in the same way. Rather, it broadly affirms existing ones, such as in relation to “non-discrimination” (Article 17), “persons with disabilities” and “children” (Article 18), and “privacy and personal data protection” (Article 11). These are more sign-posts than directives.

Nevertheless, such an agreement does have some implications – for example: 

First, the Convention provides a tool for human rights advocates to advance arguments about transparency, digital literacy, and participation. It is an agreement which has a pro-democratic ethos, which is not to be sniffed at, not least at a time when democracy, rights, and the rule of law are all under threat across the globe.

Second, it situates the Council of Europe as something of a forum for discussion on such issues. Although more impactful discussions may take place in Brussels, Washington DC, and (perhaps) London, the potential for the Council of Europe to convene state and non-state actors from across and beyond Europe is welcome.

Third, although the Convention does not establish directly enforceable rights, or establish any judicial mechanisms, the European Court of Human Rights has a significant history of interpretation of the ECHR consistent with other international treaties. The same could occur in relation to the AI Convention. That would, admittedly, require Article 21 (the Convention shall not be construed as “limiting, derogating from or otherwise affecting the human rights or other legal rights and obligations…”) to act as a floor rather than a ceiling. Time will tell.