The UN Human Rights Council has adopted a Poland-led resolution setting out safeguards for the use of artificial intelligence by governments and public institutions, including requirements for human oversight, transparency and opportunities for people to challenge AI-assisted decisions. The resolution, adopted in Geneva without a vote on 6 October, also calls for human-rights assessments before and after potentially sensitive systems are deployed.
Officially titled The role of good governance in the promotion and protection of human rights, the measure is Human Rights Council Resolution 63/12. The Council’s official resolution record shows that it was submitted by a core group consisting of Poland, Australia, Chile, South Korea and South Africa and was adopted as orally revised without a vote.
Poland has led work on the Council’s recurring good-governance resolution since initiating it in 2008. The Polish Foreign Ministry says this year’s version was ultimately co-sponsored by 58 countries from different regions of the world.
- Find more news from Poland on our Homepage .
Warsaw described the document as the first of its kind in the Council’s history to establish principles specifically governing responsible use of AI by public institutions from a human-rights perspective. That description should be understood in the context of earlier work: the Council’s 2024 good-governance resolution had already addressed AI and commissioned a study into its impact, but the new text goes substantially further by spelling out detailed safeguards for its use in public administration.
The UN text submitted to the Council calls on states to ensure that AI systems used in public administration comply with international human-rights law throughout their life cycle, from design and procurement to deployment, operation and eventual decommissioning.
Among the most significant provisions is a call for governments to consider human-rights due diligence and impact assessments before deploying systems and periodically afterwards. Those assessments should, where relevant, examine effects on women, people with disabilities and children.
The resolution also calls for human oversight and legal accountability when AI contributes to public decisions. Authorities should provide sufficient transparency and explainability for AI-supported decisions, while affected people should have a meaningful opportunity to understand and challenge them. Independent oversight should examine issues including discrimination, accuracy and compliance with human-rights law, with effective remedies available where violations occur.
The safeguards extend into particularly sensitive areas of government activity. AI-assisted surveillance, including facial recognition and other biometric identification technologies, should comply with principles including legality, necessity and proportionality and be subject to effective independent oversight.
For courts and prisons, the text says introducing AI must preserve human oversight, judicial independence, procedural fairness and human dignity. Decisions affecting a person’s liberty should not be based solely on automated systems.
The resolution also addresses a less technical but potentially important consequence of digitalising government services. States are urged to maintain non-digital ways of accessing public services for people who cannot, or do not wish to, use digital channels, while ensuring AI systems do not reproduce or worsen discrimination.
Companies supplying AI systems to governments are included as well. The text calls on businesses involved in designing or providing public-sector AI to carry out human-rights due diligence, maintain appropriate documentation and risk controls, provide safeguards and grievance mechanisms and address harmful human-rights effects associated with their systems.
The Council has also asked the Office of the UN High Commissioner for Human Rights to turn the principles into practical guidance covering procurement, deployment and use of AI across public administration. The work is expected to address impact assessments, discrimination, independent oversight, auditing and access to remedies before being presented to the Council at its 69th session.
The resolution does not create a new binding international AI law. Human Rights Council decisions are not legally binding on states, although the UN describes them as instruments that can establish standards and provide guidance to governments.
Its significance instead lies in creating an internationally endorsed human-rights framework for a field in which governments are rapidly experimenting with automated systems for public services, administrative decisions, justice and surveillance.
The initiative also gives Poland a substantive role in shaping that debate. During preparations for the Council’s 63rd session, Poland said it was presenting the good-governance resolution on behalf of the five-country core group, while Polish representatives argued that the rapid spread of AI through public institutions was becoming one of the major issues requiring international attention.
For citizens, the central principle is comparatively straightforward: using an algorithm does not remove a government’s responsibility for the decision it makes. Under the framework endorsed in Geneva, public-sector AI should remain subject to human accountability, scrutiny and the same fundamental rights that apply when decisions are taken without automated technology.
