UNESCO adopts a recommendation on the ethics of AI
In November 2021 the 193 member states of UNESCO adopted the Recommendation on the Ethics of Artificial Intelligence. The 43-page document rests on four values and ten principles, among them proportionality and do no harm, the right to privacy and data protection, and safety and security, and adds concrete areas of policy action from gender to data to international co-operation.
Why it matters
This is the broadest text on AI by membership that the atlas holds: where the OECD recommendation gathered its own circle of states and six partners, here every member of the organisation signed. It is no more binding than that one; what matters is that a shared wording proved possible outside a single economic bloc.
The four values in UNESCO's phrasing: respecting and promoting human dignity and human rights; fostering just, peaceful and interconnected societies; ensuring diversity and inclusiveness; and supporting the flourishing of the environment and ecosystems. There are ten principles; the first three are proportionality and do no harm, meaning that use must not go beyond what is necessary for a legitimate aim and that risk assessment should prevent harms; safety and security, meaning that both unwanted harms and vulnerabilities to attack should be addressed; and the right to privacy and data protection. The document code is SHS/BIO/PI/2021/1 and it is licensed Creative Commons BY-NC-SA. The record gives no day of adoption: UNESCO's own page names only the month, and the catalogue record gives 2022 as the year of publication. Nor does it claim that any binding norm followed from this text. The phrase first global standard on AI ethics is UNESCO's and is reported as its claim rather than as the atlas's judgement.