Whilst the EU’s AI Act has been grabbing all the headlines, the draft AI Convention has been flying under the radar. But not for much longer.
The Convention, also sometimes called the ‘AI Treaty’, has been drafted by the Council of Europe. If adopted, it will be the first legally binding international convention on AI. The latest draft is not the final version, but contains text that is still subject to negotiation.
We’ve set out our top five talking points on the draft Convention below:
- Who does the Convention apply to? It will apply to all signatories to the Convention which are expected to include most of the Council of Europe’s 46 member states, in particular EU Member States and the UK. Each party will need to adopt measures (including legislation where appropriate) to give effect to the Convention.
- Scope of the Convention: Whilst it is clear that the Convention applies to AI systems that have the potential to interfere with human rights, democracy, and the rule of law, some aspects of its scope are still subject to debate. This includes whether the Convention should apply to research and development, or to systems that protect essential national security interests (such as foreign intelligence and counter-intelligence activities).
- Common principles: Chapter III sets out general common principles that each party will need to incorporate in the lifecycle of AI systems in a manner appropriate to its domestic legal system. These principles will be familiar to those already working on AI governance projects, and include: human dignity and individual autonomy; transparency and oversight; accountability and responsibility; equality and non-discrimination; privacy and personal data protection; preservation of health and the environment; reliability and trust; and safe innovation.
- Remedies: Each party will need to take measures to ensure the availability of accessible and effective remedies for violations of human rights resulting from the activities within the lifecycle of AI systems. This includes measures to ensure that certain AI systems are properly documented, and that appropriate information is recorded, provided to bodies authorised in with domestic law to access that information and, where appropriate, made available or communicated to affected persons.
- What’s next? This latest draft reflects the second reading of the draft Convention on AI. It will serve as the basis for the third and final reading (date to be confirmed).
There are many questions that are yet to be debated further or defined, for example, around the synergy between this Convention and other AI legislation, notably the EU AI Act. There also appears to be more discussion to be had around the scope, with some commentary that the Convention is too wide, yet others suggest it should be narrowed, for example, by not applying to public sector organisations. So, lots to still discuss and clarify however definitely a case of ‘watch this space’!