Noting that what repositories qualify as AI projects changes over time. This from the caption on the data page: "Caution is advised when comparing different versions of the data, as the AI-related concepts identified by the machine learning algorithm may evolve in time." They're chosen based on a machine learning algorithm:

OECD.AI partners with GitHub to identify public AI projects – or “repositories” – following the methodology developed by Gonzalez et al. (2020). Using the 439 topic labels identified by Gonzalez et al. – as well as the topics “machine learning”, “deep learning”, and “artificial intelligence” – GitHub provides OECD.AI with a list of public projects containing AI code. GitHub updates the list of public AI projects on a quarterly basis, which allows OECD.AI to capture trends in AI software development over time. [1]

I'm not sure but this makes it sound like the whole time series changes quarterly when they refresh/decide anew what "concepts" qualify as AI?

I might quit this question..... May look at Gonzalez et al paper first, which OECD doesn't properly cite, but which I think it's safe to assume is this: DOI:10.1145/3379597.3387473 from MSR '20: 17th International Conference on Mining Software Repositories.

Also confusing: Looking at just the number (not percentage) of contributions (commits), I see 14.19865521 commits in 2021 for EU27?? How are there fractions of commits? What am I missing?

[1] Methodology page: https://oecd.ai/en/github