Of Artificial Intelligence and Fundamental Rights Charters

In the past five years, the Council of Europe has made significant strides to address the urgent need for a dedicated instrument that protects democracy and fundamental rights in the context of disruptive technologies and AI. Acting with remarkable speed, it has adopted the Framework Convention on Artificial Intelligence and Human Rights, Democracy, and the Rule of Law – the first of its kind. Notably, the Framework Convention includes provisions specifically tailored to enable the European Union’s participation. At the same time, the EU has developed its own complex legal framework around AI, striking a careful balance between technological advancement and human rights.

Together, these legal instruments hold the potential to strengthen the safeguarding of fundamental rights in Europe in an era defined by rapid technological advancements. Yet, the diversity of these legal sources also contributes to a complex and fragmented landscape. To better harmonize these frameworks and safeguard democracy and fundamental rights from technological misuse, I argue that the EU should adopt the Framework Convention, making an essential first step toward integrating the protection of fundamental rights of the EU Charter of Fundamental Rights and of the European Convention on Human Rights. Ultimately, this should help to create a common constitutional language where national apex courts retain the independence to establish their own national standards while consistently referencing both the EU legal framework, particularly the Charter, and the Framework Convention.

The possibility of the EU accession to the Framework Convention

The EU signed the Framework Convention on 5 September 2024. The Framework Convention is unique in many ways (including being the first AI treaty), but another singular aspect is the relationship between the Framework Convention and the EU. The Framework Convention, which is currently signed by 7 State members of the Council of Europe as well as 2 States that are non-Members, provides for the EU to join with the status of sui generis organisation that characterises the autonomy of the EU legal order (see on this Lenaerts, 2018; Nic Shuibhne, 2019; Lionello, 2024). Provided it is not challenged by EU Member States before the Court of Justice of the EU, this accession will mark the first time the EU joins a Council of Europe convention.

Many will recall the ongoing saga of the accession of the EU to the European Convention of Human Rights (ECHR), where the Court of Justice eventually determined that the 2013 Accession Agreement “is liable adversely to affect the specific characteristics of EU law and its autonomy” (para 200, Opinion 2/13 of 18 December 2014). It should be noted that the Framework Convention does not mention directly the ECHR, and perhaps rightfully so. The process of EU accession to the Framework Convention and the parallel process of EU accession to the ECHR should be kept clearly apart. However, the difference lies in the fact that the Framework Convention has been designed with the EU accession in mind, including a specific norm aimed at preserving the autonomy of the EU legal order. Article 27, paragraph 2, of the Framework Convention allows the Member States to continue applying EU law rules on AI within the EU internal market, provided that it does not affect the full application of the Convention (paras 147 and 148 of the Explanatory Report to the Framework Convention).

The Council of Europe Framework Convention and the EU legal framework

The Council of Europe Framework Convention obviously does not operate in a legal vacuum. The legal framework on AI in the EU is already fairly advanced, with Regulation (EU) 2024/1689 (the AI Act) being the key piece of legislation. Additionally, AI is going to be regulated by other instruments which are currently being negotiated, including the revision of the Product Liability Directive and the Directive on Non-contractual Liability. Other EU legal acts that are already in force and will inevitably impact on AI are, of course, the General Data Protection Regulation (GDPR), the Digital Services Act (DSA), the Media Freedom Act and the Platform Workers Directive. The GDPR is already mentioned several times in the AI Act, while the DSA and the Media Freedom Act have specific provisions that will also be applicable to AI products that operate in the field of digital services (such as social media) or media freedom (as AI that can be used to produce or fabricate news contents). The Platform Workers Directive contains rules on the algorithmic management of workers that will also be applicable to enterprises and businesses using artificial intelligence.

The AI Act, in particular, has the potential to be closely intertwined with the Framework Convention, as it was negotiated concurrently and will serve as the primary instrument for its implementation in the event of the EU’s accession to the Framework Convention (on the differences between the AI Act and the Framework Convention see Ziller, 2024). Article 27 of the AI Act mandates a fundamental rights impact assessment for AI high risk systems that should align with the impact assessment outlined in the Framework Convention. Consequently, the Framework Convention will serve as a crucial instrument for EU judges, enabling them to interpret the impact assessment instruments in the AI Act and in other EU secondary legislation in a manner that protects democracy, fundamental rights, and the rule of law.

The Framework Convention and the Charter of Fundamental Rights of the EU

The main question is how the Framework Convention will interact with other instruments for the protection of fundamental rights in Europe, and in particular with the Charter of Fundamental Rights of the EU.

As mentioned above, Article 27 of the Framework Convention explicitly allows the application of EU rules on AI among EU Member States. The Charter is applicable to the EU institutions and to the Member States when implementing EU law (Article 51). This means that once EU law applies, the Charter is applicable as well. The Charter is almost 18 years old, considering its latest proclamation in Strasbourg in 2007, and almost 25 years old if we take into account its proclamation in Nice in 2000. Henceforth, one could argue that the Charter needs a revision to effectively interact with the Framework Convention and the EU body of secondary legislation on AI.

However, I contend that such an amendment is unnecessary. In fact, the Charter is already well-suited for the digital age. Specifically, the fundamental rights of the first generation, outlined in Titles I and II of the Charter, are applicable to situations involving artificial intelligence, as they result from the implementation of EU law or the work of EU institutions. A strictly positivist approach (the tension between EU law and legal positivism has been described masterfully on this Blog by La Torre, 2024) might suggest that the Charter requires amendment because the legal issues arising from AI and disruptive technologies were not considered during its drafting. Yet, the counterargument is that the Charter should be interpreted as a living instrument by both EU and national judges (Palmisano, 2014).

Additionally, it is often argued that the Charter is not applied extensively by national judges in domestic disputes, and this can be justified for several reasons. Some judges may defer to higher courts that, depending on national procedural autonomy, could directly or indirectly discourage lower courts from applying primary EU law. Others, although this is becoming less common, might refrain from applying primary EU law (and the Charter) due to the complex case law governing its scope of application. Finally, much to the dismay of European law scholars, a significant number of national disputes lack a clear link with EU law.

To promote its application, the EU should continue to support the dissemination of the Charter through targeted funding and proactive initiatives that demonstrate how to effectively utilize existing fundamental rights instruments to safeguard democracy and fundamental rights in the context of AI and other disruptive technologies.

Once these measures are in place, the Framework Convention is more likely to serve as a valuable tool for national judges, enabling them to interpret other EU legal instruments – particularly the Charter – in ways that address situations where democracy, fundamental rights, and the rule of law are at stake.

Embracing complexity but avoiding overcrowding

The opportunity lies in the potential of the Framework Convention to assist judges in clarifying the application of EU law to situations where democracy and fundamental rights intersect with artificial intelligence, bridging the EU and the Council of Europe legal systems. Although the tide has been partially turned during the trialogues on the AI Act, it is difficult to overlook that the AI regulation in the EU is primarily designed around the internal market – and perhaps rightfully so. Therefore, once EU accession to the Framework Convention is finalized, it may aid both national and EU judges in accurately interpreting and applying the EU framework on AI.

To add complexity to the picture, this delicate role played by the Framework Convention must be balanced with that of national constitutional courts, which often compete with other European courts to ensure the most appropriate degree of protection. However, the problem lies in the overcrowding of charters of rights and instruments for the protection of fundamental rights. The exponential increase in the number of instruments and levels of protection could ultimately undermine legal certainty.

One suggestion is to foster a common constitutional language of digital fundamental rights in Europe by incentivising last-instance national courts and constitutional courts to reference the Charter, as well as the Framework Convention, and other international conventions alongside national constitutional or primary law provisions. Ultimately, interpretation of national and EU legal instruments will rest with the competent court – whether EU, international or national – but this approach might help navigate the complexities arising from the stratification of legal instruments while preserving legal certainty.