UN rights office calls for stronger human rights checks on AI systems
An advance unedited version of an OHCHR report says states must carry out continuous human rights due diligence across the AI lifecycle, especially in high-risk areas such as policing, welfare, migration and military use.
The Office of the UN High Commissioner for Human Rights has published an advance unedited version of a report on human rights due diligence for AI.
The report was submitted to the Human Rights Council’s 62nd session. It focuses on how states should meet their obligations to respect, protect and fulfil human rights when AI systems are developed, procured, deployed or regulated.
OHCHR says AI can create risks that are systemic, large-scale and difficult to detect. These include discrimination, unlawful surveillance, privacy violations, automated denial of public services, biassed decision-making and harms linked to military or security uses.
The report argues that states cannot treat AI risk assessments as one-off exercises. Human rights due diligence should be continuous and should cover the full AI lifecycle, from design and training data to deployment, monitoring and remedies.
It also calls for stronger safeguards in rights-sensitive areas. These include law enforcement, border management, migration, welfare administration and military applications. Where serious risks cannot be effectively reduced, OHCHR says states should refrain from using certain AI systems or prohibit them altogether.
The report also highlights public procurement as a major gap. Governments often buy or use AI systems developed by private companies, but procurement rules may not include strong human rights standards, transparency or accountability.
OHCHR recommends public registries of AI systems used by governments, independent human rights impact assessments, meaningful human oversight and the right to challenge AI-assisted decisions.
The report also says companies across the AI supply chain, including developers, data brokers, infrastructure providers and deployers, should be subject to effective regulation and human rights due diligence requirements.
OHCHR concludes that AI governance must be based on transparency, public participation, independent oversight and access to remedy.
