The UK government has officially signed the first-ever “legally binding” treaty concerning artificial intelligence (AI) and human rights, committing signatory states to implement safeguards against various risks associated with AI technology.
Formulated by the Council of Europe, an organization established in 1949 to promote human rights across Europe, the treaty has garnered signatures from Andorra, Georgia, Iceland, Norway, the Republic of Moldova, San Marino, as well as Israel, the United States, and the European Union (EU).
Entitled the “Framework Convention on Artificial Intelligence, Human Rights, Democracy, and the Rule of Law,” the treaty delineates several key principles that states must follow throughout the entire lifecycle of AI systems. These principles encompass privacy and data protection, transparency and oversight, equality and non-discrimination, safe innovation, and the preservation of human dignity.
To safeguard these principles, the treaty mandates that countries establish measures to evaluate and mitigate any potentially harmful impacts of AI, alongside providing effective remedies for any human rights violations that may occur as a result of AI operations.
“AI has the potential to significantly enhance the responsiveness and efficiency of public services and drive economic growth,” stated Lord Chancellor and Justice Secretary Shabana Mahmood. “However, it is crucial that we direct AI’s development rather than allowing it to dictate terms to us. This convention represents a significant advance in ensuring that these emerging technologies are utilized without undermining our fundamental values, such as human rights and the rule of law.”
Peter Kyle, Secretary of State for Science, Innovation, and Technology, emphasized that the treaty will play a vital role in unlocking the potential of AI to enhance economic growth and public services: “Once enacted, it will bolster protections for human rights, the rule of law, and democracy, reinforcing our domestic stance on technology while advancing the global agenda for safe, secure, and responsible AI.”
While the treaty applies to AI in the public sector—including instances where private companies act on behalf of the government—it does not specifically address private sector AI use. Instead, it leaves it up to individual states to determine the applicability of the treaty’s requirements to private enterprises.
Additionally, the treaty includes a specific exemption for national security interests. It states that “a Party shall not be required to apply this Convention to activities within the lifecycle of artificial intelligence systems related to the protection of its national security interests, provided such activities comply with applicable international law.”
Though a member state can prohibit certain AI applications deemed incompatible with human rights, the treaty does not specify penalties for government violations.
Some experts have pointed out that the treaty’s compliance mechanisms are somewhat ambiguous, with only a requirement for periodic reporting on efforts to meet its standards. “There are no strict enforcement criteria, which may limit the treaty’s effectiveness and impact,” noted lawyers from Bird&Bird. However, the treaty does include mechanisms for resolving disputes between governments regarding its implementation and allows countries to “denounce” (i.e., withdraw from) the convention by notifying the Secretary-General of the Council of Europe.
Nick Reiners, a senior geo-technology analyst at Eurasia Group, remarked to Gzero Media that the treaty’s voluntary nature reduces its legal bindingness, despite claims from its signatories. He also highlighted that the national security exemption could dilute its rigor, citing instances like Israel’s use of AI in Gaza for targeting, which would remain unaffected.
Reiners noted that the EU’s participation is likely part of an effort to “internationalize the AI Act,” encouraging non-European nations to align with its technological priorities.
The treaty will take effect three months after it receives ratification from at least five signatories, including three from the Council of Europe, allowing governments worldwide to join thereafter.