The OECD Council adopts a recommendation on AI
On 22 May 2019 the Council of the Organisation for Economic Co-operation and Development adopted Recommendation OECD/LEGAL/0449: five values-based principles for responsible stewardship of trustworthy AI and five recommendations to governments. Beyond the member countries, Argentina, Brazil, Colombia, Costa Rica, Peru and Romania adhered to it. The OECD described these at the time as the first such principles signed up to by governments.
Why it matters
This is the first intergovernmental text in which requirements on AI systems are put as a norm rather than as an industry's appeal to itself. It is not binding, and the OECD says so plainly. What matters here is the form: after it, transparency, robustness and accountability could be discussed in words that states had already signed.
The five principles in the OECD's own 2019 phrasing: inclusive growth, sustainable development and well-being; respect for the rule of law, human rights, democratic values and diversity, with appropriate safeguards such as enabling human intervention; transparency and responsible disclosure so that people understand AI-based outcomes and can challenge them; robust, secure and safe functioning throughout the life cycle with risks continually assessed; and accountability for those who develop, deploy or operate the systems. The five recommendations to governments cover investment in research, accessible ecosystems, an enabling policy environment, skills and support for workers through the transition, and cross-border co-operation. The record gives no total count of adhering states: the page that names them gives no number. Nor does it claim that any other text grew out of this one. The text now standing on the OECD's legal instruments site is the version as amended on 3 May 2024, where the principles are worded differently. The descriptions above are taken from the OECD's own page as archived on 1 June 2019, ten days after adoption.