The OECD.AI Policy Observatory is a database of global AI polices and AI policy research was launched on February 27th, 2020 at the OECD Conference in Paris, France. In this post I provide an overview of what the Observatory is, and discuss why I think the "Business Stakeholder Initiatives" section which houses AI Ethics Policies for 19 leaders in AI is a great resource for organizations developing their own AI ethics policies.
For those that need a quick refresher: OECD is the Organisation for Economic Co-operation and Development, an intergovernmental organization with 36 member countries (soon to be 37 with the forthcoming addition of Colombia) with a mandate to shape global policies to foster prosperity, equality, opportunity and well-being for all. Country membership includes the UK, US, Canada, Mexico, Australia, New Zealand, Korea, Japan, Israel, Turkey, Chile, as well as the majority of EU member states. This group was behind the OECD Principles on AI launched in May 2019, the same principles that inspired the G20 AI Principles (more on the principles below) launched later in June 2019.
Onto the Observatory!
The new OECD.AI Policy Observatory, launched in February 2020 is a brain-child of the OECD Recommendation on AI released in Spring 2019. The Observatory's aim, as stated on the site is to "help countries encourage, nurture and monitor the responsible development of trustworthy artificial intelligence (AI) systems for the benefit of society." Its shining star in my opinion is the Countries & Initiatives section - a database of global AI policies and AI policy research from governments, non-government partners, researchers, and other parties interested in AI policy. The OECD hopes that cataloging AI policy resources will help facilitate dialogue and provide multidisciplinary, evidence-based policy analysis on AI.
Within the Countries & Initiatives section, there are two primary areas: stakeholder initiatives and national AI policies.
The stakeholder initiatives section houses without a doubt the most helpful page for organizations - the Business Stakeholder Initiatives page, one of six stakeholder initiative pages (the others are academia, technical communities, civil society, intergovernmental organizations, and trade unions). Need to build an AI Ethics Code of Conduct? Interested in what other organizations are doing? This page can help as it houses 19 AI Ethics Codes of Conduct (at the time of publishing on March 5th) of the leading AI companies across various industries and countries. The usual suspects are present with Google, IBM, Microsoft all present, but the observatory team has gathered documents from Baidu, Telefonica, SONY, Deutsche Telekom, and Sage along with several other organizations to provide the largest list of organizational AI ethics initiatives I've seen by far.
The national AI policies section offers users the option to search AI Policies by country or territory, type of policy instrument used, or by the group targeted by the policy. Searching by country or territory will lead you to 59 different country policies, plus those of the EU; by far the most comprehensive list I've seen (previously the most comprehensive overview covered 25 countries plus the EU - an early and widely-referenced article by Tim Dutton).
Each country or territory has its own "dashboard" housing (Check out the example for Canada here):
A list of the country's AI initiatives, with a summary page for each as well as a link to the original document
An overview of the policy instruments used by category
Country specific related news
Policy instruments organized by number and budget
A list of organizations in the country responsible for each policy
The groups targeted by the policies
Budgets
Related OECD publications, and
A list of the top AI research institutions.
Although some of the detailed data may be overkill for a regular organization, the list of the country's AI initiatives and the corresponding overview pages is incredible helpful for those trying to get a grasp on a country's AI policies.
The Countries & initiatives section is however just one of the four pillars of OECD.AI Policy Observatory, the four pillars (as presented along the website's top menu) are as follows:
1) AI Principles: Which provide an overview of the OECD Values-based Principles on AI, the Recommendations for policy makers, as well as important AI terms & concepts
2) Policy Areas: Which houses the OCED-led AI policies and policy research organized by field (e.g. finance and insurance, education, environment, agriculture). Each field has it's own page with related publications, online news, and recent scientific research, as well as several visualizations of data on the policies. If you're interested in the OECD AI policy for a given field then this section may be helpful, but with very limited external AI policies or research for each field, it's likely not going to be helpful for organizations.
3) Trends & Data: A series of reports on the methodological work done by the OECD to quantify AI developments around the globe, along with several visualizations of data on AI across countries, institutions, topics, and policy areas. Could be useful if you're an organization looking for some country-specific data on AI growth, but again, likely the Countries & Initiatives section will be more valuable.
4) Countries & Initiatives***: As discussed, this is the shining star of the observatory for organizations as it houses several hundred AI policies in an easily searchable database, with great summaries of the national initiatives.
Overall, the OECD.AI Policy Observatory is a huge step forward in aligning the AI ethics and policy vision around the globe. As the OECD was able to corral opinions with the launch of its AI Principles back in 2019, I have hopes that the OECD.AI Policy Observatory will be able to do the same for AI policy globally moving forward. For organizations, it's a valuable resource, and I hope to see more organizations adding to the list of Business initiatives in the coming months.
You can visit the homepage for the OECD.AI Policy Observatory here: https://oecd.ai/
Comments