UN rights council adopts safeguards for artificial intelligence

The UN Human Rights Council adopted its first artificial intelligence resolution by consensus on October 7, 2026. The decision in Geneva places human rights at the centre of how AI systems are designed, deployed and monitored, giving governments a common political framework as the technology enters public administration, workplaces and daily life.

The resolution says AI must comply with international law at every stage, from manufacture to use. It also calls for a ban on applications that fail to respect human rights. The language covers the full lifecycle of systems, rather than focusing only on the moment a tool reaches the public.

For Switzerland, the decision carries particular weight. The country is a leading AI developer and Geneva hosts many of the institutions shaping international digital policy. Swiss companies, public authorities and researchers will face the same questions as governments elsewhere: who is responsible when an automated system causes harm, and how can affected people challenge a decision?

The resolution is not a treaty and does not create an enforcement mechanism. Its importance lies in the principles it puts on the record, including legality, accountability and protection from discrimination. The Council is due to return to the subject next June.

Human oversight must remain in place whenever AI influences a decision, according to the resolution. States are urged to ensure that a person supervises decisions involving AI and that responsibilities are clearly assigned. That requirement targets a recurring weakness in automated systems: institutions can rely on a tool without making clear who reviews its output or answers for the consequences.

The Council also calls on governments to assess human rights impacts when they introduce laws and regulatory frameworks. Authorities should address discrimination and other harmful effects, including risks that arise from data, design choices or the way systems are used in practice.

The wording has direct implications for public services. An algorithm used in welfare administration, policing, migration processing or recruitment can affect access to rights and opportunities even when no official describes the process as fully automated. Human review will need to be meaningful, with enough information and authority to question or overturn a system’s recommendation.

The resolution leaves implementation to states. Switzerland’s federal authorities, cantonal administrations and municipalities therefore retain responsibility for translating the principles into procurement rules, impact assessments and complaint procedures. The practical test will be whether people can understand and contest an AI-assisted decision.

The Council wants every country, including developing nations, involved in setting AI rules. The resolution argues that global governance cannot be shaped only by the states and companies with the greatest computing power. It calls for stronger international cooperation and greater coherence between the many initiatives now under way.

That appeal reflects a practical divide. Governments need expertise, computing capacity and regulatory resources to assess advanced systems, while communities affected by automated decisions may have limited access to technical knowledge or legal remedies. Broad participation is intended to make standards more legitimate and reduce the risk that rules reflect only the priorities of wealthy countries and major technology firms.

The resolution also asks states, the United Nations and other stakeholders to promote digital public goods, particularly by improving access to digital infrastructure. That provision links human rights protection with the ability to connect, use digital services and participate in public debate.

Geneva will remain a focal point for this discussion. The next Global Summit on AI is scheduled to take place there in June. Switzerland’s diplomatic role, technical institutions and international organisations give the country a platform, but also create expectations that its own policies will support open participation and rights-based safeguards.

The resolution arrives as Switzerland weighs the opportunities and risks of being a leading AI developer. Swiss universities, companies and public institutions are building and using systems that can process information, automate tasks and influence decisions. The country’s role gives it an interest in clear international expectations, especially where systems operate across borders.

The UN text does not prescribe a single regulatory model. It asks states to consider human rights when they legislate, prevent discrimination and establish clear responsibility. For Switzerland, that points toward coordination between federal authorities, cantons, regulators, researchers and the private sector.

Different cantons may encounter AI in different settings and at different speeds. A municipal service, a hospital, a school or a police department may purchase a system from a supplier operating outside Switzerland. Common standards for documentation, testing, oversight and redress could help public bodies evaluate those tools before they affect residents.

The Council’s approach also recognises that regulation must keep pace with technical change. The Office of the High Commissioner for Human Rights will submit a report within one year, providing a further opportunity to examine how the principles work in practice. Swiss officials and organisations will be watching closely as the debate moves from Geneva’s diplomatic language to operational rules.

The Council adopted a second resolution focused on women and girls in the digital environment. It urges states to protect their human rights and prevent gender-based violence facilitated by digital technologies. The measure places online abuse and technology-enabled harm within the same broader human rights discussion as automated decision-making.

The two resolutions address different risks, but both call for governments to take responsibility for the effects of technology. AI systems can reproduce discriminatory patterns, while digital platforms and other tools can be used to threaten, target or silence women and girls. Prevention therefore requires more than technical performance checks. It also requires effective laws, institutional accountability and access to remedies.

The Council adopted both measures by consensus, giving them a shared political mandate even though implementation remains national. Switzerland will need to consider these principles across its digital policy, including education, law enforcement, public services and support for people facing technology-facilitated abuse.

The immediate next steps are institutional. The Council plans further discussion in June, and the UN human rights office has a one-year reporting deadline. Those processes will show whether states can turn broad safeguards into standards that public authorities, technology companies and individuals can use when AI and digital tools affect real rights.