First International AI Treaty Adopted by Council of Europe
The Council of Europe has adopted the first-ever international treaty on artificial intelligence, establishing a legal framework for transparency, accountability, and human rights. The move complements other global initiatives, such as the European Parliament's proposed copyright changes for generative AI and declarations from Cannes and Delhi promoting AI sovereignty and ethics.
- The treaty provides signatory nations with two compliance options for the private sector: they can either be directly bound by the convention's provisions or implement "other appropriate measures" to meet the treaty's principles, a flexibility introduced after the U.S. advocated for an alternative to direct application. - For companies operating within the European Union, the treaty's principles will be implemented through the existing EU AI Act, which classifies AI systems into risk tiers and carries significant financial penalties for non-compliance, with fines up to €35 million or 7% of global annual turnover for prohibited systems. - Enforcement of the treaty itself does not include fines but relies on a "Conference of the Parties" to monitor implementation and ensure compliance among signatory nations, establishing an international standard for accountability. - The convention requires signatories to establish legal frameworks for risk and impact assessments throughout the AI lifecycle, a provision directly applicable to the deployment of agentic and autonomous systems, especially in high-risk sectors like finance and healthcare. - Developers and deployers of AI are required to ensure transparency, such as notifying individuals when they are interacting with an AI system, and to provide effective remedies for those harmed by an AI system's decisions. - The framework is designed to be "technology-neutral," focusing on the application and impact of AI systems rather than specific technologies, aiming to remain relevant as AI capabilities, including autonomous workflows, evolve. - While the treaty covers the entire lifecycle of an AI system, it includes specific exemptions for matters of national security and for activities purely related to research and development, unless testing has the potential to interfere with human rights or democracy. - This agreement creates a legally binding international baseline that complements other global AI governance efforts, contrasting with national regulations like the U.S. Executive Order on AI and setting a global standard grounded in human rights principles.
