The Special Rapporteur’s 2018 report to the General Assembly is now online. 

As reliance on Artificial Intelligence (“AI”) grows in today’s information environment, it is critical to address its impact on the exercise of freedom of expression and related human rights. On one hand, AI may enable broader, quicker and more targeted means of sharing information and ideas globally. On the other, the opacity of AI systems deployed by both State and non-State actors may also interfere with the formation and development of opinions, access to information and other fundamental aspects of individual autonomy and agency.

This report examines key human rights concerns raised by major applications of AI, particularly in the areas of web search, content curation, content moderation and commercial advertising. Drawing on the UN Guiding Principles on Business and Human Rights and past reporting, it also outlines a rights-based framework for the design, procurement and operation of AI systems. When governments rely on AI systems to execute core public functions, they must ensure that their design and operation comply with international human rights standards through a rigorous cycle of public consultations, due diligence and audits. As part of their duty to promote freedom of expression, governments should also ensure and support rights-oriented approaches to AI in the private sector, including measures to ensure a diverse, pluralistic and competitive information environment.

Companies are also responsible for ensuring that AI and related technologies deployed on their platforms or sold to third parties are consistent with human rights standards. Ethics-based frameworks are not enough. Instead, the design and deployment of AI to generate, collect and analyze information about end users must treat international standards on freedom of opinion and expression as their authoritative reference point. A rights-oriented approach also demands human rights impact assessments, transparency measures, regular independent and external audits and consultations with affected end users and other stakeholders. In particular, these processes should prioritize the need to counter and address the persistent threat of discrimination and bias in the design and deployment of AI.