Elaborating on the first binding international framework for artificial intelligence, the Council of Europe, an organization that represents up to 46 European countries, approved this Friday, after two years of negotiations, a landmark treaty that binds states to regulate AI. While welcomed as a milestone, the move also sparked debate among experts and policymakers who question its scope and practical impact.
A key provision of the convention obliges signatories to implement measures to identify, assess, prevent, and mitigate potential risks associated with AI. It also sets out how these technologies should be controlled to ensure they do not undermine democratic institutions and processes, and to guarantee respect for equality, including gender equality, prevent discrimination, and protect private life. The pact frames a preventive approach grounded in human rights principles and tailored to the realities of modern governance and public life.
Ahead of the formal signing on September 5, signatory countries will be free to adopt the already established provisions or to activate other comparable measures. The framework thus offers immediate pathways for adoption and potential room for future enhancements as technology evolves and new challenges emerge.
The Council of Europe, which operates independently of the European Union, comprises 46 European states. It counts six additional countries—the United States, Canada, Japan, Mexico, Israel, and the Vatican—as observers. Argentina, Australia, Costa Rica, Peru, and Uruguay have also participated in the negotiations, signaling broad attention to AI governance beyond the European continent.
Private Sector and Police
Despite the two-year internal deliberations that shaped the treaty, the agreement has stirred dissatisfaction among AI experts and lawmakers alike. Critics argue that the duties designed to safeguard human rights do not fully apply to national security contexts. The deployment of systems such as facial recognition in policing or at border checkpoints remains especially controversial because it may intensify discrimination against vulnerable communities, including migrants. The regulation framework of the EU itself carries that contentious exception, which has fed ongoing debate about what protections are feasible in security-focused scenarios.
On April 18, the Parliamentary Assembly of the Council of Europe issued a statement condemning the agreement for largely excluding the private sector, which drives much of AI development. This differentiated approach, they argued, creates a significant gap. Lawmakers also proposed adding explicit provisions addressing health and environmental considerations to strengthen the framework and broaden its scope to real-world impact across sectors.
Recently, on Wednesday, Volker Türk, the United Nations High Commissioner for Human Rights, urged the Council of Europe to rethink these two controversial exemptions. For journalist Luca Betuzzi, these critiques signal a clear mismatch between the authorities’ triumphalist narrative and the lived reality of AI deployment, suggesting a need for more balanced governance that can gain public trust and practical legitimacy.
Overall, the agreement represents a careful attempt to balance the opportunity AI offers with the responsibilities it imposes. It highlights the ongoing tension between safeguarding fundamental rights and enabling innovation, a balance that will require ongoing dialogue among policymakers, industry, civil society, and affected communities. The conversations will likely shape future amendments and complementary measures that respond to emerging use cases and the rapid pace of technical change, ensuring that AI serves the public good while reducing harm and bias.[citation]n[attribution]n