Historic first-ever AI treaty

On 14 March, an agreement was reached on the Council of Europe's Artificial Intelligence, Human Rights, Democracy and the Rule of Law Framework Convention, the first international treaty on AI.

On the occasion of the finalisation of the Convention, Secretary General Marija Pej?inovi? Buri? stated:

'This first-of-a-kind treaty will ensure that the rise of Artificial Intelligence upholds Council of Europe legal standards in human rights, democracy and the rule of law. Its finalisation by our Committee on Artificial Intelligence (CAI) is an extraordinary achievement and should be celebrated as such'

'[The treaty] sets out a legal framework that covers AI systems throughout their lifecycles, from start to end.'

'The text strikes the right regulatory balance precisely because it has benefitted from the input of governments and experts, and industry and civil society. We thank all of those partners for their contribution and delivering this seminal text. We are convinced that, once adopted, this treaty will bring everyone together in appreciation of its impact.'

Delving Deeper into the Convention

Bilateral negotiations shape scope

The negotiation process, primarily conducted between the European Commission and the US delegation, focused intensely on defining the treaty's scope. The main cause of friction was that the US tried to exclude the private sector from the AI treaty. Notably, a plenary meeting was convened late in the evening to address contentious issues.

Scope and coverage

The Convention encompasses activities throughout the AI system lifecycle that possess the potential to impact human rights, democracy, and the rule of law. However, it applies exclusively to actions undertaken by public entities and private actors acting on their behalf.

Addressing risks from private actors

Participating countries are obliged to manage risks and impacts arising from private actors not affiliated with the public sector in line with the treaty's objectives. Upon signing or ratifying the Convention, countries must declare their intentions regarding the application of obligations to private actors or adopt alternative measures.

Flexibility in convention

Importantly, countries retain the flexibility to amend their declarations at any time. This allows for adjustments to commitments, giving nations the ability to adapt their stance as needed.

Limitations on national security

While the treaty extends coverage to various aspects of AI governance, certain limitations exist concerning national security interests. Countries are not mandated to apply the treaty to activities related to safeguarding national security. However, such activities must align with international law and democratic processes.

Criticisms

Despite its historic significance, some view the treaty as falling short of its intended impact. Concerns have been raised regarding the treaty's effectiveness, with suggestions that it primarily reaffirms existing practices rather than introducing substantive regulatory measures. Just a few days ago, the EU data watchdog expressed concerns about potential compromises in human rights standards due to pressure from foreign business interests. The EDPS believes that the proposal does not go far enough in addressing the risks and challenges posed by AI. The text has been significantly weakened from its original version during negotiations at the CoE's ad hoc committee in charge of the convention. The EDPS described it as a 'missed opportunity to lay down a strong and effective legal framework' for protecting human rights in AI development.

What is next?

According to the preliminary timeline, the convention will be put to the vote at the ministerial-level meeting in May 2024.