The
Framework Convention on Artificial Intelligence, drawn up by the Council of Europe and signed by the United States, United Kingdom and the European Union marks a significant step towards mitigating the risks posed by AI while promoting innovation.
At its core there is the mandate that AI systems must uphold fundamental human rights and democratic values, ensuring their deployment does not lead to discrimination or violations of privacy. This lifecycle approach covers the entire process, from the development of AI technologies to their eventual deployment and operation, integrating ethical considerations at every stage.
According to the Council of Europe, the goal of the newly signed Framework Convention on Artificial Intelligence is to "fill any legal gaps that may result from rapid technological advances" in AI. Council of Europe Secretary-General Marija Pejcinovic Buric emphasized the importance of ensuring "that the rise of AI upholds our standards, rather than undermining them." The Council of Europe, designed the treaty to "promote AI progress and innovation, while managing the risks it may pose to human rights, democracy and the rule of law."
Governments are expected to introduce a range of safeguards, such as stemming the spread of AI-generated misinformation and preventing AI systems from being trained on biased data, which could lead to wrongful decisions in areas like job applications or benefit allocations.
Until now, there has been a series of regulations and agreements covering AI, from the EU's AI Act to last year's Bletchley Declaration at the inaugural Global AI Safety Summit, as well as a voluntary testing regime signed by numerous countries and companies. This new international treaty represents an attempt to create a cohesive global framework.
Entities utilizing AI systems are required to explicitly inform users when they are interacting with an AI system rather than a human – a crucial measure to prevent deception. The treaty calls for concerted efforts to combat the spread of false and misleading information, aligning with growing concerns about the potential for AI to be exploited for the dissemination of propaganda.
Its scope extends beyond the public sector, applying to both governmental institutions and private organizations. The European Union's involvement is particularly significant, as the treaty complements the bloc's recently enacted AI Act, which regulates the use of AI within the internal market.
In addition to the EU, US, and UK, the treaty was also signed by Andorra, Georgia, Iceland, Norway, Moldova, San Marino, and Israel. Several other countries, including Argentina, Australia, Canada, and Japan, were involved in the negotiations.
The Council of Europe has described the treaty as an "open" agreement with the potential for global adoption. The organization has urged more countries to sign and ratify the framework, which it hopes will create a comprehensive international governance system for AI technologies.
Concerns have been raised about the treaty's exemptions for national security applications, which could potentially be interpreted broadly by signatory countries.
Francesca Fanucci from the European Center for Not-for-Profit Law told Reuters that the "formulation of principles and obligations in this convention is so overbroad and fraught with caveats that it raises serious questions about their legal certainty and effective enforceability."
Fanucci also pointed to exemptions for national security applications and limited scrutiny of private companies compared to the public sector, which could undermine the treaty's overall effectiveness.
Additionally, the treaty's reliance on individual countries to implement its provisions through domestic laws and regulations introduces a level of uncertainty regarding enforcement and compliance. The lack of a centralized oversight mechanism, akin to the European AI Act's European AI Office, may result in varying degrees of implementation and enforcement across the signatory nations.
While the AI Convention is an international treaty with the potential for broader global adoption, the EU AI Act is a regulation that applies directly to all 27 member states of the European Union and has extraterritorial reach, covering any AI system used within the EU.
Additionally, the AI Convention allows for more flexibility in exemptions, particularly related to national security applications, compared to the more limited exemptions in the EU AI Act.